Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
SHRINK FIT EFFECTS ON ROTORDYNAMIC STABILITY: EXPERIMENTAL
AND THEORETICAL STUDY
A Dissertation
by
SYED MUHAMMAD MOHSIN JAFRI
Submitted to the Office of Graduate Studies of Texas A&M University
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
May 2007
Major Subject: Mechanical Engineering
SHRINK FIT EFFECTS ON ROTORDYNAMIC STABILITY: EXPERIMENTAL
AND THEORETICAL STUDY
A Dissertation
by
SYED MUHAMMAD MOHSIN JAFRI
Submitted to the Office of Graduate Studies of Texas A&M University
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
Approved by: Chair of Committee, John M. Vance Committee Members, Alan B. Palazzolo Luciana R. Barroso Guy Battle Head of Department, Dennis L. O’Neal
May 2007
Major Subject: Mechanical Engineering
iii
ABSTRACT
Shrink Fit Effects on Rotordynamic Stability: Experimental and
Theoretical Study. (May 2007)
Syed Muhammad Mohsin Jafri, B.E., NED University of Engineering &
Technology, Karachi;
M.S., Texas A&M University
Chair of Advisory Committee: Dr. John M. Vance
This dissertation presents an experimental and theoretical study of sub-
synchronous rotordynamic instability in rotors caused by interference and shrink fit
interfaces. The experimental studies show the presence of strong unstable sub-
synchronous vibrations in two different rotor setups with interference and shrink fit
interfaces that were operated above their first critical speeds. The unstable vibrations
occur at the first natural frequency of the rotor-bearing system. The instability caused
complete wreckage of the test rig in one of the setups showing that these vibrations are
potentially dangerous to the safe operation of rotating machines. The two different rotor
setups that are studied are a single-disk rotor mounted on a uniform diameter shaft and a
two-disk rotor with an aluminum sleeve shrink fitted to it at the outer surface of the two
disks. In the single-disk rotor, an adjustable interference arrangement between the disk
and the shaft is obtained through a tapered sleeve arrangement, which acts as the
interference fit joint. The unstable sub-synchronous vibrations originate from slippage
in the shrink fit and the interference fit interfaces that develop friction forces, which act
as destabilizing cross-coupled moments when the rotor is operated above its first critical
speed. The unique contribution offered through this work is the experimental validation
of a physically correct model of internal friction which models the destabilizing
iv
mechanism as a system of cross-coupled internal moments at the shrink fit interface. The
dissertation describes stability simulations of various test rotor setups using the correct
internal moments model. A commercial finite-element based software called XLTRCTM
is used to perform rotordynamic simulations for stability studies. The method of stability
study is the computation of eigenvalues of the rotor-bearing system. A negative real part
of the eigenvalue indicates instability. The simulations include the test rotors that were
experimentally observed as stable and unstable with shrink and interference fit interfaces
in their assemblies. The dissertation also describes the simulations of various imagined
rotor configurations with shrink fit interfaces, and seeks to explain how configurations
differ on rotordynamic stability depending upon several rotor-bearing parameters such as
geometry and elastic properties, as well as upon the amount of internal friction
parameters, which differ from configuration to configuration.
v
DEDICATION
To Almighty Allah for His help and blessings
To Abbu, Ammee, Deeju and all of my Family
“Study thou, in the name of thy Lord who created;- Created man from Clots of Blood:-
Study thou! For thy Lord is the most Beneficent Who hath taught the use of the Pen;-
Hath taught Man that which he knoweth not…”
Al-Koran- Chapter XCVI The Clot
vi
ACKNOWLEDGEMENTS
My first and foremost acknowledgement in completing this intellectually
challenging project that culminated in my Doctorate dissertation goes to Dr. John M.
Vance. I am grateful to him for providing me with an opportunity to work on the project,
along the way providing me unconditional technical and moral support. I have benefited
immensely from his vast and extremely useful knowledge while completing my
dissertation. I will forever remember him as being one of the greatest influences of the
development of my technical knowledge and insight into the highest level of
engineering.
My sincere appreciation is for my committee members, Dr. Palazzolo, Dr. Battle
and Dr. Barroso, for agreeing to serve on my committee. I took classes under Dr.
Palazzolo and Dr. Battle and benefited immensely from them, not only learning new
material from the classes, but actually applying the knowledge and skills gained from
there for the completion of my dissertation.
My sincere appreciation for the Turbomachinery Research Consortium (TRC) for
their financial support of the Internal Friction Project. Without their support, it would not
have been possible to conduct any experiments and therefore, the research would never
have materialized.
Last, but not least, my acknowledgements go to my colleagues and co-workers at
the Turbomachinery Laboratory, especially Eddie Denk, for helping me tremendously
along the way in my research. Without Eddie’s help and guidance at the machine shop
and the test cell, no substantial progress would have been possible in these experiments.
I do not have enough words to thank Eddie for his selfless help.
I am thankful to all of you.
vii
NOMENCLATURE
Z,Y,X Inertial frame coordinate axes
z,y,x Rotating frame coordinate axes
t Time [T]
ω Rotational speed of a rotor [1/T]
Ω Precessional or whirling speed of a rotor [1/T]
θ Angular micro-slip at a shrink fit interface about the X-axis [-]
φ Angular micro-slip at a shrink fit interface about the Y-axis [-]
fθ Forward whirl component of micro-slip about the X-axis [-]
bθ Forward whirl component of micro-slip about the X-axis [-]
α Angular micro-slip at a shrink fit interface about the x-axis [-]
β Angular micro-slip at a shrink fit interface about the y-axis [-]
)(•
Differentiation with respect to time [( ) / T]
θθK Direct moment stiffness at a shrink fit interface about the X-axis [FL]
φφK Direct moment stiffness at a shrink fit interface about the Y-axis [FL]
θφK Cross-coupled moment stiffness about the X-axis [FL]
φθK Cross-coupled moment stiffness about the Y-axis [FL]
θθC Direct moment damping at a shrink fit interface about the X-axis [FL]
φφC Direct moment damping at a shrink fit interface about the Y-axis [FL]
θM Moment at a shrink fit interface about the X-axis [FL]
φM Moment at a shrink fit interface about the Y-axis [FL]
αM Moment at a shrink fit interface about the x-axis [FL]
βM Moment at a shrink fit interface about the y-axis [FL]
i Imaginary number operator ( 1− ) [-] tie Ω Complex exponential-harmonic function [-]
viii
sgn The signum function ( 1± ) [-]
dissE Energy dissipated [LF]
E Modulus of elasticity of a solid [F/L2]
r Radial coordinate [L]
σ Applied stress on a solid [F/L2]
ν Poisson’s ratio value of a solid [-]
N Normal reaction or force from contact [F]
rσ Radial stress at an interface [F/L2]
tσ Tangential stress at an interface [F/L2]
0δ Radial interference or shrink fit at zero rotational speed [L]
)(ωδ Radial interference or shrink fit as a function of rotational speed [L]
ψ Circumferential location of a point at an interface [-]
R Interface radius [L]
L Axial contact length of an interface [L]
slidingV Relative sliding velocity at an interface [L/T]
re Unit vector in radial direction as measured in x,y,z frame [-]
ψe Unit vector in circumferential direction as measured in x,y,z frame [-]
Sμ Coefficient of static friction [-]
Kμ Coefficient of dynamic friction [-]
ix
TABLE OF CONTENTS
Page
ABSTRACT ..................................................................................................................... iii
DEDICATION ...................................................................................................................v
ACKNOWLEDGEMENTS ..............................................................................................vi
NOMENCLATURE.........................................................................................................vii
TABLE OF CONTENTS ..................................................................................................ix
LIST OF FIGURES...........................................................................................................xi
LIST OF TABLES ..........................................................................................................xiv
CHAPTER
I INTRODUCTION: THE IMPORTANCE OF THE RESEARCH………….1
Background of problem…………………………………………………..1 Literature review…………………………………………………………3 Dissertation objectives…………………………………………………..10 Research methodology…………………………………………………..10
II EXPERIMENTAL TEST FACILITY...........................................................12
Drive motors…………………………………………………………….12 Instrumentation………………………………………………………….13 Test rotors……………………………………………………………….15 Stiffener structures………………………………………………………18 Experimental results…………………………………………………….19 Formula for calculating the radial interference fit………………………23
III INTERNAL FRICTION MOMENTS MODEL ...........................................34
Gunter's follower force model…………………………………………..35 Internal moments model………………………………………………...40
IV EQUATIONS OF CROSS COUPLED MOMENTS FOR THREE INTERFACE FRICTION MODELS………………………........................45
Basic rotordynamic model for analysis…………………………………47 Kinematics of rotor motion……………………………………………..52 Physical interpretation of friction models………………………………64
x
CHAPTER Page
V EXPLANATION OF KIMBALL’S MEASUREMENTS USING INTERNAL MOMENTS MODEL.................................................................76
Basic theory of rotor internal friction…………………………………...76 Modification to Kimball's hypothesis…………………………………...78
VI ROTORDYNAMIC MODELING USING XLTRCTM ..................................83
Overview of modeling using XLTRCTM………………………………..84 Construction of a single-disk rotor model………………………………87
VII ROTORDYNAMIC SIMULATIONS OF EXPERIMENTS USING THE INTERNAL MOMENTS MODEL........................................................97
Single-disk rotor simulations……………………………………………98 Two-disk rotor simulations…………………………………………….107
VIII CONCLUSIONS ..........................................................................................115
Recommendations for future research…………………………………116
REFERENCES...............................................................................................................118
APPENDIX A ................................................................................................................121
APPENDIX B ................................................................................................................143
APPENDIX C …………………………………………………………………………149
APPENDIX D ………………………………………………………………………....185
APPENDIX E………………………………………………………………………….195
VITA………………………………………………………………………………….. 230
xi
LIST OF FIGURES
FIGURE Page
1 Drive motor arrangement with belt and bearings supporting the drive shaft.......12
2 Double-row self aligning ball bearing used with the bearing housing for rotor support .........................................................................................................13
3 A Metrix proximity probe ....................................................................................14
4 Proximitors and power supply for powering the proximity probes .....................14
5 Close-up view of a single-disk rotor bearing system tested at the Turbomachinery Laboratory ................................................................................15
6 Single-disk rotor installed on the ball bearings at the Turbomachinery Laboratory ............................................................................................................16
7 Two-disk rotor installed on the ball bearings at the Turbomachinery Laboratory ............................................................................................................17
8 Another view of the two-disk rotor, showing the steel rotor disk which is shrink fitted with the aluminum sleeve at the ends ..............................................17
9 Single-disk rotor on foundation............................................................................19
10 Two-disk rotor on foundation ..............................................................................20
11 Shrink fit in the single-disk rotor due to tapered sleeve ......................................22
12 Positions of draw bolts and push bolts on tapered sleeve ...................................22
13 Waterfall plot of test 1 showing significant instability starting from 5800 rpm .............................................................................................................25
14 Bode plot of test 1 showing growing amplitudes of vibrations above 5800 rpm ..............................................................................................................26
15 Waterfall plot of test 2..........................................................................................27
16 Bode plot of test 2 ................................................................................................28
17 Waterfall plot showing the threshold speed at 11,000 rpm..................................29
18 Waterfall plot showing threshold speed of instability at 9600 rpm .....................31
19 Spectrum plot from LVTRC showing the sub-synchronous instability component ............................................................................................................32
xii
FIGURE Page
20 Wrecked two-disk rotor.........................................................................................33
21 Extended Jeffcott rotor model .............................................................................35
22 Cross-section of the extended Jeffcott rotor showing the moment and force vectors .........................................................................................................36
23 Free-body diagram of a rotor with internal friction, according to Gunter ..........37
24 Modeling of internal friction using Gunter’s model ............................................39
25 A two-disk rotor whirling in first mode, along with a stiff aluminum sleeve....................................................................................................................40
26 Free-body diagram of the shaft carrying steel wheels..........................................41
27 Free-body diagram of the shaft showing equivalent internal friction moment vectors ....................................................................................................43
28 Free-body diagram of the shaft showing equivalent internal friction moment vectors ....................................................................................................44
29 Sketch of a flexible vibrating shaft with a shrink fitted sleeve. ...........................47
30 Side view of the rotor model from Fig.29, showing the discontinuity of slope at the shrink fit interface between the shaft and the sleeve.........................48
31 A model of the shrink fit interface friction showing the spring elements.. ..........50
32 Coordinate transformation between the fixed and the rotating frames of reference ...............................................................................................................51
33 End view of the disk showing the radial stresses acting on its surface ................56
34 Kelvin-Voigt model of internal friction in solids.................................................65
35 Hysteresis loop due to viscous friction ................................................................66
36 Schematic description of a shrink fit interface using a torsional spring and a damper ........................................................................................................68
37 A mechanical model for illustration of the hysteretic friction .............................69
38 Hysteresis loop for hysteretic friction model .......................................................70
39 Mechanical model to explain Coulomb friction...................................................72
40 A magnified view of the irregularities at the mating surfaces .............................73
41 Friction force F as a function of applied force P..................................................74
42 A rotor disk supported on a flexible shaft rotating clockwise..............................76
xiii
FIGURE Page
43 Rotor disk side views (a) Purely elastic deflection (b) Deflection with internal friction.....................................................................................................77
44 Side view of the disk-shaft under the influence of internal hysteresis.................79
45 Frictional tensile and compressive stresses acting on the shaft ...........................80
46 Top-view of the rotor showing the bent centerline due to friction moments .......81
47 Finite element model of a single-disk rotor using XLTRCTM..............................87
48 Rotor model with bearing connections, connecting the rotor with the ground...................................................................................................................89
49 Rotor model with bearings and foundation included ...........................................91
50 Interface points of Shafts 1 and 2, where internal friction parameters are specified.. .............................................................................................................93
51 Internal moments acting on the rotor disks (in its plane of deflection) in X and Y directions and their resultant moment vector, MR .....................................94
52 XLUseMoM Worksheet for entering of internal friction parameters ...................96
53 Close-up view of the single-disk rotor showing the shaft and disk interface through tapered sleeve...........................................................................98
54 Single-disk rotor simulation using XLTRCTM .....................................................99
55 Unstable mode shape of the single-disk rotor above the threshold speed..........103
56 Unstable mode shape of the single-disk rotor above the threshold speed (for the case of tight fit)......................................................................................106
57 Isometric view of the two-disk rotor showing internal features.........................107
58 Unstable mode shape of two-disk rotor model with different fits at two interfaces ............................................................................................................111
xiv
LIST OF TABLES
TABLE Page
1 Coefficients of the internal moment applied at the interface of the shaft and the disk for loose fit. ....................................................................................101
2 Coefficients of the internal moment applied at the interface of the shaft the shaft for tight fit............................................................................................105
3 Coefficients of internal moment applied at the undercut end for two-disk rotor simulation ...................................................................................109
4 Coefficients of internal moment applied at the tight fit end for two-disk rotor simulation. ..................................................................................110
5 Coefficients of the internal moment applied at the two interfaces for the same fit case of the two-disk model .......................................................113
1
CHAPTER I
INTRODUCTION: THE IMPORTANCE OF THE RESEARCH
BACKGROUND OF PROBLEM
In the early 1920’s it was observed that, with some rotors running well above
their first critical speed, there occurred a series of rotor wrecks and damages which at
first were not understandable and were attributed to improper balancing of the rotors.
The General Electric Company (GE) encountered a series of serious damages to their
blast furnace compressors running well above their first critical speeds. Dr. B.L.
Newkirk from the GE Research Laboratories was appointed to research and investigate
these damages and come up with some practical solutions to these problems. Newkirk
discovered oil whip from fluid-film bearings as one of the causes of these rotors wrecks.
However, there were other rotors operating without the fluid-film bearings and they had
similar wreckages. At this time (1924), Dr. A.L. Kimball from the GE came up with an
explanation of the latter type of rotor behaviors and proposed internal friction as a cause
of rotor damages. He maintained that during rotational motion of the rotors, the rotor
shafts bend and produce longitudinal friction forces inside the rotor material itself. This
friction produces a disturbing torque on the rotor shaft, causing the shaft to move in the
forward whirl direction, when the rotational speed is above the first critical speed. When
the rotational speed is below the first critical speed of the rotor, the internal forces tend
to dampen the system and reduce the vibrations. However, above the first critical speed,
these forces provide a positive energy input to the system and thus increase the
vibrations level, leading to the rotor damage.
This recognition of damping acting as energy addition to the system as in
contrast to the strictly accepted view of damping as an energy dissipation was a
remarkable intellectual achievement. It continues to be an intellectually challenging
This dissertation follows the style and format of Journal of Applied Mechanics.
2
problem to understand as to how a damping which is produced within a rotating system
itself can lead to destabilization of the rotor motion, and in most cases, can cause serious
rotor wreckages. Kimball showed in his paper, by deriving the equations of motion that
the internal friction force tends to put the shaft motion in an ever-increasing spiral path.
In the language of vibrations theory, this is called rotordynamic instability.
Thus, the phenomena came to be known as rotordynamic instability due to
internal friction. He also modeled the internal friction force due to shrink fits in the rotor
systems and mentioned that the effects of internal friction due to shrink fits are far more
pronounced and predominant than those due to internal friction in the rotating shaft
itself. Newkirk confirmed Kimball’s observations through experiments and proposed
shrink fits as the main reason for this rotordynamic instability.
Today, internal friction is seen to be a potential source of rotordynamic problems
in advanced, high pressure Oxygen-Hydrogen propulsion equipment [1]. Turbopumps
such as the Space Shuttle Main Engine (SSME) High Pressure Oxidizer Turbopump
(HPOTP) are of built-up design with many joints, fits, and areas for friction-induced
excitation if slippage takes place. These rotors operate at supercritical speeds with light
external damping. The power densities of these turbomachines are high. Therefore, the
forces on the rotors are very large, which tends to encourage joint slippage and friction
force generation. This has resulted in highly expensive and troublesome shut downs of
machine operations at various leading turbomachinery users such as National
Aeronautics Space Administration (NASA) and General Electric (GE) [1, 2], to name
only but a few. Therefore, the importance of research underlies in the motivation to
safeguard the expensive rotating machines against permanent and costly damages and to
understand the mechanics of destabilizing forces and moments produced due to slippage
in shrink fit and interference fit interface joints, so as to propose better designs to the
industry that will ensure stable operations throughout the operating speed range of
turbomachines.
3
LITERATURE REVIEW
The design philosophy applied to rotating machinery initially began with the
construction of very stiff rotors that would ensure operation below the first critical
speed. It was only after Jeffcott’s [3] analysis in 1919, when he showed that the rotors
could be operated safely beyond their first critical speeds with proper rotor balancing
that the trend in rotordynamics design changed. As the rigid rotor model was replaced by
more flexible models, several failures were encountered when operating at speeds above
the first critical speed. Most of the failures were of unknown origin at that time. Newkirk
[2] of the General Electric Research Laboratory investigated the failures of compressor
units in 1924, and found that these units encountered violent whirling at speeds above
their first critical speeds, with the whirling rate equal to the first natural frequency. If the
rotor speed were increased above its initial whirl speed, the whirl amplitude would
increase, leading to the rotor failure. The speed at which the rotor begins to whirl is the
threshold speed of instability. Kimball [4], working with Newkirk, suggested the internal
friction as a cause of shaft whirling. He showed that below the first critical speed, the
internal friction would damp out the whirl motion, while above the first critical speed, it
would sustain the whirl.
After a series of experiments on internal friction, Newkirk and Kimball arrived at
a number of conclusions, the most important being: (1) the onset speed of whirling and
the whirl amplitude is unaffected by the rotor balance, (2) whirling always occurs above
the first critical, (3) whirling is encountered only in the built-up rotors, and (4)
increasing the foundation flexibility or increasing the damping to the foundation
increases the whirl threshold speed.
Gunter [5] explained some of the experimental results of Newkirk. He developed
a linear rotordynamic model which includes the effects of bearings and foundation
support flexibility and damping, besides the flexibility and internal damping of the rotor.
He modeled the internal friction as a cross-coupled force. Through this model, he
showed that external damping stabilizes the rotor bearing system, by increasing its
threshold speed of instability. However, there is a limit to the external damping; a so-
4
called optimum damping that stabilizes the rotor. He also showed that the foundation
flexibility, even without external damping, stabilizes the rotor. This means that no
additional external damping is required to stabilize an unstable rotor; support flexibility
alone may prevent a rotor to become unstable. However, in the case of fluid-film
bearings, in which there is an appreciable amount of the cross-coupling forces due to
thin, pressurized oil films which support the rotor loads, there is a strong tendency for
the oil whirl and whip, in which case it is necessary to have an external damping source
to stabilize the rotor running above its first critical speed.
Walton, Martin, and Lund [1, 6] conducted experimental and theoretical research
on internal friction using a test rotor facility with axial spline and interference (shrink) fit
joints. They proposed the internal friction model as a system of internal moments rather
than the forces. Transient and steady-state simulations of their internal friction model
showed close agreement with the experiments on their test rotor. Their experiments
showed that both the axial spline joint and the shrink fit joints cause sub-synchronous
instabilities and in some cases, super-synchronous instabilities at the rotor’s first natural
frequency. Their experiments also showed that the dry-film lubrication in the axial joints
causes the instability component. Balancing of the rotor does not decrease the sub-
synchronous instability due to the shrink fits to the same extent as the synchronous
component is decreased. They modeled the rotor using finite elements and employed a
Coulomb friction model to analyze both the axial spline joints and the shrink fit joints.
Kimball [7, 8] described experimental measurements of internal friction in
different rotor materials, both with and without shrink fits. He postulated that internal
hysteresis in a material during spin will cause the shaft to deflect sideways in the
direction of forward whirl. He measured the magnitude of internal friction force by the
sideways deflection of a loaded overhung shaft during spin. From the measurements, he
concluded that the sideways deflection is independent of the spin velocity (or the rate of
strain of the shaft fibers) and that shrink fits cause larger deflection of the shaft as
compared to the case of no shrink fit on the shaft. These experiments showed that shrink
5
fits, rather than the material internal hysteresis, are a much more important mechanism
of the forward whirl instability.
Lund [9] analyzed various models of internal friction due to axial splines and
shrink fit joints in a rotor. His analysis showed that cross-coupled moments developed
due to internal friction at the interface of the joints are a cause for rotor instability.
Specifically, his analysis showed that the linear viscous damping model predicts
instability above the first critical speed of the rotor, subject to the condition that the
external backward whirl stabilizing effect due to bearing support asymmetry should not
exceed the forward whirl destabilizing effect, whereas the solid friction model predicts
some instability ranges above the first critical speed. He also showed that a micro-slip
model for the axial splines predicts rotor instability above certain whirl amplitudes when
the rotor speed exceeds the first critical speed.
Artilles [10] analyzed the effects of internal friction on rotor stability due to axial
spline couplings. In the analysis, the internal friction is modeled as a system of cross-
coupled moments which are developed at the spline interface due to relative sliding
between the spline teeth. The simulations of the non-linear differential equations of
motion for a rotor model which include the cross-coupled moments highlight the effects
of various system parameters such as unbalance, side loads and initial conditions on the
stability of the rotor. The simulations showed that the amplitude of the unstable sub-
synchronous component is not dependent on the amount of imbalance included in the
model. In most cases of the simulations, limit cycle amplitudes are predicted for the sub-
synchronous component.
Black [11] analyzed different internal friction models for investigating the
stability of a flexible rotor supported on damped, flexible bearings with no cross-
coupling. The internal friction models were viscous friction, Coulomb friction and
hysteretic friction. He showed that the viscous friction model predicts a threshold speed
of instability for the rotor-bearing system which is greater than the rotor first critical
speed, with the value of the threshold speed of instability dependent on external and
internal damping parameters of the system. The analysis of the viscous friction model
6
predicted the rotor instability once the threshold speed is reached. For the Coulomb
friction model, Black’s analysis predicted that the rotor-bearing system becomes
unstable as soon as the first critical speed of the rotor is traversed, if a certain parameter
of the rotor, called relaxation strength (an indicative of friction inside the shaft material)
is greater than twice the external damping ratio. This model predicts instability above the
first critical speed for all subsequent higher speeds, subject to the condition that the
relaxation strength is higher than the external damping ratio for instability to occur. That
is, if the relaxation strength is not greater than twice the external damping ratio, the
operations above the first critical speed will be stable, according to the Coulomb friction
model. The hysteretic friction model predicts a range of speeds (above the first critical
speed) in which the rotor-bearing system becomes unstable, but above that range, the
operation is stable. The hysteretic friction and Coulomb friction are more realistic
models as compared to the viscous friction model, due to: (1) prediction for a range of
limited unstable operation, (2) instability of the rotor upon traversing its first critical
speed. Both of these predictions have been verified experimentally. Black also analyzed
the effects of bearing stiffness asymmetry on the rotor stability and concluded that the
stiffness asymmetry promotes the system stability while the damping asymmetry
demotes the stability to some extent.
Ehrich [12] presented a model of internal friction which showed that the internal
friction stresses act in a direction perpendicular to the shaft deflection plane and that
their magnitude is proportional to the rate of change of strain of the shaft fibers. His
analysis showed that the ratio of the threshold speed of instability to the first critical
speed depends upon the amount of internal and external damping of the rotor. His
analysis also predicts instability above multiple critical speeds and shows that it is not
necessarily the first mode of the rotor-bearing system which is always excited in an
unstable whirl caused by the internal friction, but that it can be any mode, including any
higher than the first mode, that can be excited.
Yamamoto and Ishida [13] formulated internal friction as a system of internal
moments, which do not produce any instability below the first critical speed, but produce
7
a sub-synchronous forward whirl instability component above the first critical speed.
Their formulation showed that the internal moments based on Coulomb’s friction model
are non-linear functions of the rotor’s instantaneous position. Their formulation is of the
same form as Walton, Martin and Lund’s [1] formulation.
Vance and Ying [14] conducted an experimental study on a two-disk steel rotor
with an aluminum sleeve having two shrink fit interfaces with the disks. Their
experiments consisted of transient (run-up and coast-down) and steady state (fixed
speed) tests of the rotor which supported Black’s analysis [10]. Some tests showed that
as soon as the rotor’s first critical speed was traversed, the forward whirl instability
appeared, suggesting the Coulomb friction model. The experiments also showed that the
instability appeared in some speed ranges, both during the run-up and the coast-down
and not over all the speeds above the first critical speed. The later observation is
predicted by the hysteretic friction model. They also utilized a heat gun in the steady-
state tests to heat the aluminum sleeve above the first critical speed and observed violent
instability of the rotor, due to loosening of the shrink fit and generation of the internal
friction effects at the shrink fit interface due to possible sliding/slipping between the
disks and the sleeve which was caused by thermal expansion of the aluminum sleeve at
the interface.
Mir [15] conducted rap tests on a single-disk rotor with an adjustable interference
fit mechanism. His experimental results showed both Coulomb and hysteretic damping
caused by interference fit in the rotor. He showed that the presence of either Coulomb or
the hysteretic damping is dependent upon the amplitude of excitation of the rotor. From
the rap test experiments, he showed that logarithmic decrement of the time traces of
rotor vibrations decreased by increasing the tightness of the fit. From the analysis of
logarithmic decrements, he concluded that the hysteretic damping coefficients will vary
with the running speed. He acquired the data for forward whirl instability caused by
interference fit in the running tests as the initial interference was reduced.
Srinivasan [16] conducted free-free tests on the same single-disk rotor as
described in Chapter II of this dissertation. From the experiments, he obtained the time
8
traces of rotor vibrations with the interference fit values varied over a certain range. By
analyzing the time traces, he obtained logarithmic decays and equivalent damping
coefficients for internal friction in the rotor (because there was almost negligible
external damping during those free-free tests). He converted the damping coefficients
into equivalent cross-coupled coefficients to model the internal friction acting in
mutually orthogonal directions. With these inputs in the XLTRCTM software, he
predicted the single-disk rotor’s threshold speed of instability, although his experimental
observations concerning the threshold speed of instability were not always repeatable.
Anand’s experimental work also showed that with a bonding tape wrapped around the
shaft, internal damping in the rotor material was enhanced as shown by logarithmic
decrements obtained from the free-free tests of the rotor.
Murphy [17] analyzed the effects of cross-coupled stiffness and damping
coefficients as well as direct stiffness coefficients on the stability of a simple rigid rotor
model, which is supported in horizontal and vertical directions by linear bearings. His
analysis showed that the direct stiffness asymmetry stabilizes the rotor, whereas the
cross-coupled stiffness coefficients cause instability when they are equal and opposite in
sign and exceed certain range of values. With equal direct stiffness coefficients with
cross-coupled stiffness coefficients of equal magnitude and opposite signs, the analysis
predicts rotordynamic instability.
Robertson [18] described the elastic hysteresis and the clamping fit effects and
how they are destabilizing to the rotors running above their first critical speeds. He
described the inadequacy of the linear viscous damping model and showed that the
internal damping can be described more accurately with the hysteretic damping model.
He also showed that just as Kimball’s and Newkirk’s explanation for the elastic
hysteresis (which depends upon the normal strain rates) results in a destabilizing force in
the direction of the forward whirl for a rotor running above the first critical speed,
similarly, a clamping fit effect such as that due to shrink fit and flexible couplings,
creates friction forces that oppose relative motion between the rotating parts. These
forces induce instability in the forward whirl for the rotors running above their first
9
critical speeds, but act as an external, stabilizing damping below the first critical speeds.
He also described and discussed some potential designs in the rotating machineries that
should be avoided to prevent rotordynamic instability caused by the internal friction.
Smalley and Pantermuehl et al. [19] presented an analysis of some centrifugal
compressor designs assembled with shrink fit joints. The analysis investigated the
stiffening effect caused by the shrink fits on the centrifugal compressors. ANSYS
software was used to investigate the compressor designs. Through the analysis, they
showed that the shrink fits stiffened the compressors and raised their first critical speeds
slightly. In some cases, the increase in the first critical speeds was as high as 6 percent.
They further showed that the increase in stiffness is proportional to the interference fit’s
length to the shaft diameter ratio. The larger this ratio is the larger is the stiffness
induced in the model. The numerical simulations from the models validate the field data.
Nelson and McVaugh [20] applied finite element analysis technique to rotor-
bearing systems. The rotor model can be a uniform or a non-uniform shaft, with any
specified number of inertias. They considered all six degrees of freedom for the model.
They demonstrated the finite element methodology and the solutions by demonstrating a
numerical example of a rotor. They incorporated internal friction in the analysis by using
two models: the linear viscous and the hysteretic. Zorzi and Nelson [21] in their analysis
included external and internal (hysteretic) damping in the equations of motion. Hashish
and Sankar [22] investigated a damped rotor system using a finite element model with
the viscous damping and the hysteretic damping as models for internal friction.
Ginsberg [23] and Meriam and Kraige [24] provide simple mechanical models of
various types of friction such as viscous friction, the Coulomb friction and the hysteretic
friction. The mechanical models used to illustrate different friction models show
important features of the models such as non-linearity in the Coulomb and the hysteretic
friction models, as well as the difference between the models such as the dependence of
friction forces on the rate of strain change in the viscous friction model and dependence
of sign of rate of strain change in the hysteretic friction model.
10
DISSERTATION OBJECTIVES
The main objectives of this research work are as follows:
1. To develop a practical capability to predict threshold speeds of whirl instability
for built-up rotors with shrink fit interfaced joints. A related goal is to develop a
way to determine the correct numerical values of the internal friction coefficients
or how cross-coupled coefficients (for any particular rotor assembly) should be
used in computer codes for stability predictions (typically logarithmic
decrement).
2. To understand how shrink fits in a given rotor assembly affect dynamic stability
of a rotor-bearing system.
The major tasks that support the main objectives (1) and (2) above are as follows:
(a) To conduct experiments with different rotors and shrink fit setups to observe
instability and establish some values of the shrink fits that induce forward whirl
instability.
(b) To develop various computer models with different geometries and
configurations for a single-disk and a two-disk rotor to understand and develop a
pattern for instability in various system parameters, such as the geometry and the
material properties, which could be related to stability of the single-disk and the
two-disk rotordynamic systems.
(c) To explain the experimental results on the stability of particular configurations of
the single-disk and the two-disk rotors at the Turbomachinery Laboratory.
RESEARCH METHODOLOGY
The research is divided into experimental and theoretical studies. The methodology
for the research in each of the two areas is described as follows:
11
1. For experimental study, vibration measurement results from a single-disk and a
two-disk rotor serve as a foundation on which some fundamental hypotheses and
assumptions about the effect of shrink fits on rotordynamic stability are based.
The experimental results on both the single-disk rotor and the two-disk rotors
show sub-synchronous instability due to the internal friction at the first
eigenvalue of the rotor-bearing system.
2. For theoretical study, modeling and simulation of the internal friction due to
shrink fits in a rotor-bearing assembly using XLTRCTM Rotordynamics Analysis
Software are carried out extensively to analyze the stability of various
configurations, both experimental as well as imagined.
12
CHAPTER II
EXPERIMENTAL TEST FACILITY
DRIVE MOTORS
The major parts of the test rig consist of a drive train and data acquisition
instrumentation. The drive system is a 30 hp variable speed motor that is connected to a
jackshaft via a toothed belt that has a speed ratio of 1 to 4.8 (Fig.1). The jackshaft is
mounted on two five-pad tilt pad bearings. The jackshaft is connected to the rotor by a
flexible coupling. The rotor is supported on two ball bearings of model number SKF
1215 K. The two bearings are double row self-aligning ball bearings. There are twenty
balls in each row and the ball diameter is 0.5 inch (12.5 mm) (Fig.2). The bearings are
lubricated by a pressurized lubrication oil system. The two bearings are mounted on
split-type SAF 515 pillow block housings.
Fig. 1 Drive motor arrangement with belt and bearings supporting the drive shaft
Motor
Fan
Belt
13
Fig. 2 Double-row self aligning ball bearing used with the bearing housing for rotor support
INSTRUMENTATION
The instrumentation consists of two 8 mm Metrix non-contact eddy current
proximity probes (Fig. 3) mounted on a probe pedestal which is bolted close to the mid-
span of the shaft. A keyphasor (which is also an 8 mm non-contact eddy current probe)
is mounted 15o from the vertical axis. The keyphasor measures the phase and the angular
speed of the shaft. The proximity probes are powered by 24 V Bently-Nevada
proximitors (Fig. 4). They are connected to a Bently-Nevada ADRE 208 data acquisition
system for acquiring and analyzing the running test data.
14
Fig. 3 A Metrix proximity probe
Fig. 4 Proximitors and power supply for powering the proximity probes
15
TEST ROTORS
Fig. 5 Close-up view of a single-disk rotor bearing system tested at the Turbomachinery
Laboratory
Fig. 5 shows a close-up view of a single-disk rotor tested at the Turbomachinery
Laboratory. The interference fit exists at the interface between the shaft and the disk
through a specially designed tapered sleeve. Fig. 6 shows a full view of the single-disk
rotor installed on the ball bearings:
Shaft
Sleeve Disk
16
Fig. 6 Single-disk rotor installed on the ball bearings at the Turbomachinery Laboratory
Fig. 7 shows the two-disk rotor tested at the Turbomachinery Laboratory. This
rotor has shrink fit contacts at the interface between the two steel disks and an aluminum
sleeve at the two ends of the sleeve. The interference axial length of the sleeve and the
wheel near the coupling end is 1 inches, whereas at the other end of the sleeve, the
contact length is 2 inches (Fig. 8). The axial width of both wheels is 2 inches.
Stiffener structures
17
Fig. 7 Two-disk rotor installed on the ball bearings at the Turbomachinery Laboratory
Fig. 8 Another view of the two-disk rotor, showing the steel rotor disk, which is shrink
fitted with the aluminum sleeve at the ends.
Areas of interference fits
Sleeve
Disk
Shaft
18
STIFFENER STRUCTURES
The stiffener structures are mounted on the foundation to increase the stiffness of
the foundation housing in the horizontal direction, thus reducing the asymmetry of the
bearing support. These structures are shown in Fig.6. Lower foundation stiffness
asymmetry reduces the stability of the system [2]. As experimental results for the
running tests on the single-disk and the two-disk rotor show, installing these structures
brought the onset speed of instability within the operating range of the rotor and made
the instability caused by internal friction a repeatable experiment.
Each stiffener structure is made of three steel I-beams welded together at the base
by a large plate that serves as the base. At the top, there is a thick steel plate (3/4 inches)
that connects the stiffener with the foundation housing, using capscrews. The stiffeners
are connected to the ground with the help of six foundation bolts passing through the
base plate to the ground.
With stiffeners mounted on the foundation, impact and shaker tests were
performed on the rig to determine the values of the modal mass, stiffness and damping
of the foundation housing in both the horizontal and the vertical direction. This estimate
is important, because these numerical values are required by XLTRC software to
perform the simulations of the system. Secondly, but equally important, these modal
parameters, especially stiffness, will provide an idea about the asymmetry of the
foundation. With the help of running tests and observing the onset speeds of instability,
it can then be seen how stiffness asymmetry will affect the onset speed of instability.
The determination of the modal parameters of the foundation is discussed in Appendix
D.
19
EXPERIMENTAL RESULTS
The experimental research with the single-disk and the two-disk rotor with the
shrink fit interfaces shows evidence of sub-synchronous instability of the rotor-bearing
system due to internal friction caused by the shrink fit joints. The two rotor setups are
shown in Fig.9 and Fig.10:
Fig. 9 Single-disk rotor on foundation
20
Fig. 10 Two-disk rotor on foundation
In the single-disk rotor, the shrink fit is created by a tapered sleeve which fits in
the inside diameter of the wheel and acts as an interface between the wheel and the shaft.
The shrink fit between the wheel and the sleeve is varied by changing the axial position
of the sleeve. The sub-synchronous instability at the first eigenvalue of the rotor-bearing
system (3000 cpm) occurred at a threshold speed of 6000 rpm, when the shrink fit at
6000 rpm was 1 mils radial. When the shrink fit at the zero speed was increased, the
instability was suppressed up to a speed of 11,000 rpm, when suddenly the sub-
synchronous instability re-appeared. From a shrink fit computation code (see Appendix
A), it was found that the shrink fit at 11000 rpm was also 1 mils radial. The experiments
with both the looser fit and the tighter fit were found to be completely repeatable.
21
Therefore, for the single-disk rotor, it was found that the 1 mils radial shrink fit is a
“critical” value, in that it causes the sub-synchronous instability to occur consistently.
For the two-disk rotor, the rotor threshold speed of instability was 9600 rpm. A
forward whirl sub-synchronous instability occurred at the first eigenvalue of the rotor-
bearing system (5000 cpm). In this rotor-bearing system, it was found that making one
end of the sleeve having a tighter fit on one of the wheels, while making the other end of
the sleeve a relatively loose fit on the other wheel destabilized the rotor-bearing system.
Heating of the looser end at a fixed speed of 9600 rpm was required for about 10
minutes. The heating was carried out using two heat guns on the loose end. The forward
whirl sub-synchronous instability occurred suddenly after heating for about 8 minutes
and then the instability started to increase in magnitude. Even as the rotor was coasted-
down, the sub-synchronous component persisted up to about 8000 rpm, when the whole
test rig wrecked, with the rotor completely damaged.
Both the single-disk and the two-disk rotors are 52.5 inches long. The shaft
material for the two-disk rotor is AISI 4340 steel, with an aluminum sleeve shrink fitted
at the two wheels. For the single-disk rotor, the material is AISI 4340, with a single-disk
having an outside diameter equal to 10 inches and an inside diameter of 2.5 inches
interference fitted with a uniform shaft through a tapered sleeve. For the two-disk rotor,
several configurations were tested by changing the geometry of the sleeve. In all but one
of the configurations, the rotor was found to be totally stable. The experimental results
are described on the following pages.
22
The single-disk rotor results
D Disk
Shaft Sleeve
Push Bolts
Draw Bolts
D Disk
Shaft Sleeve
Push Bolts
Draw Bolts
Fig. 11 Shrink fit in the single-disk rotor due to tapered sleeve [16]
R 1.25"
2
1
3
4
5
6
Fig. 12 Positions of draw bolts and push bolts on tapered sleeve [16]
23
In this experimental setup, a tapered steel sleeve fits inside the tapered bore of
the steel disk by elastic deformation of the sleeve, as shown in Fig.12, thereby creating
an interference fit between the disk and the shaft. There are six draw bolts and an equal
number of push bolts, that can be mounted on the holes bored at the periphery of the
tapered sleeve. These bolts are used to pull the disk up on the taper, thereby varying the
distance between the sleeve’s outer edge and the disk, and providing a way to vary the
interference fit between the disk and the shaft. The sleeve-disk schematic is shown in
Fig.11. The end view of the wheel with the sleeve and the shaft, along with the push and
the draw bolts, is shown in Fig.12.
FORMULA FOR CALCULATING THE RADIAL INTERFERENCE FIT
Based on the geometry of the sleeve-disk interface as shown in figure 8, the
following equation can be used to obtain the radial interference fit that is developed
between the disk and the sleeve:
)(2
DSTR −=δ (1)
In equation (1), ‘T’ is the taper ratio of sleeve, which is 1:24. ‘S’ is a calibration
factor, which denotes the distance between the outer edge of the sleeve and the outer
edge of the disk at zero interference fit, whereas ‘D’ is the distance between the outer
edges of sleeve and the disk that can be varied by using push and draw bolts.
From experiments on the sleeve, it is found that:
S = 1.596 in. (40.538 mm)
Hence, equation (1) can be recast as follows:
)596.1(02083.0 DR −=δ (2)
24
1. Run up tests: Speed limited to around 8500 rpm
Ying [14] showed that incipience of the forward whirl sub-synchronous
instability due to internal friction depends on tightness of the fit (not too tight). Some
previous tests also showed that the internal friction instability is neither predominant at
either too loose a fit nor at too tight a fit, but rather at some intermediate range of the
fits. The earlier experimental results on the single-disk rotor showed that the rotor-
bearing system was showing some sub-synchronous component at a fixed frequency
beginning at the running speed of around 5500 rpm, but the amplitudes were very small.
Thus, to differentiate those sub-synchronous components from any possible benign sub-
synchronous components, it was decided to conduct running tests with different
interference fits. Initially, in order to assess the instability of the system, the rotor was
run up to a speed of about 6500 rpm and then coasted down. It was observed during
those initial running tests that there was some sub-synchronous component of the
vibration with the frequency equal to the first eigenvalue of the rotor in the vertical
direction. Moreover, its amplitude was growing with every increment in the rotating
speed. The running speed was increased to about 8500 rpm in subsequent experiments
and the data was collected. Three such tests are described below:
25
(a) Test 1
Fig. 13 Waterfall plot of test 1 showing significant instability starting from 5800 rpm
Figs. 13 and 14 are the snapshots of the data acquired from the ADRE data
acquisition software. The waterfall plots in Fig.11 show a large sub-synchronous
component at a frequency near the first critical speed of the rotor, which is around 2900
rpm. As shown in Fig.13, the sub-synchronous amplitudes grow with the rotating speed
of the rotor. These are shown as red-colored lines in Figs. 13 and 14. The Bode plots in
Fig. 14 show that the instability is roughly growing with the speed of the machine with
large amplitudes (around 30 mils, peak-to-peak) at around 8000 rpm. Also, in the
rotating speed range of 6000 to 8000 rpm, it can be noticed that the 1X component is
very small, but the direct vibration component is large, showing that the instability is the
predominant component of the rotor vibration.
Sub-synchronous Component (Instability)
1X
26
Fig. 14 Bode plot of test 1 showing growing amplitudes of vibrations above 5800 rpm
27
(b) Test 2
With the same shrink fit condition, the next experiment was conducted to assess
the repeatability of the instability. The waterfall and the Bode plots for this test are
shown in Figs. 15 and 16:
Fig. 15 Waterfall plot of test 2
Again, a significant instability is observed at the same speed and at the same
frequency as in the first case (test 1). Therefore, the tests are repeatable and consistent.
28
Fig. 16 Bode plot of test 2
The initial shrink fit values for both the tests were identical and the shrink fit at
the threshold speed that caused the instability to occur was around 1 Mil (radial). This
value of the shrink fit at the threshold speed was estimated using a code for the shrink fit
variation with the rotational speed at the sleeve and the disk interface.
29
(c) Test 3
Some tests were conducted using a tighter initial shrink fit and it was observed
that the instability was suppressed at the previous threshold speed of 6000 rpm. Instead,
the threshold speed for tighter fits became 11000 rpm, with higher instability amplitude
as shown in Fig.17 below:
Fig. 17 Waterfall plot showing the threshold speed at 11,000 rpm
The two-disk rotor results
Several configurations for the two-disk rotor were tested. These are briefly
outlined below:
(1) An aluminum sleeve 9.5 inch outside diameter, with the diametral shrink fit at
both ends equal to 11 mils
(2) An aluminum sleeve 9.25 inch outside diameter, with the diametral shrink fit at
both ends equal to 7 mils
30
(3) An aluminum sleeve 10 inch outside diameter, with the diametral shrink fit at
one end being 11 mils diametral, whereas at the other end it was equal to 5 mils
diametral. The end with the 5 mils diametral interference had only 1 inch axial
contact with the corresponding steel wheel (it did not have complete 2 inch axial
contact; the sleeve was undercut at the loose end intentionally).
From the experiments, it was found that the first two configurations were
perfectly stable under all the operating conditions, and although there was some sub-
synchronous component at the first natural frequency of the rotor for the rotor spin
speeds above the first critical speed, the amplitudes of those sub-synchronous vibrations
were too small to be conclusive. The configuration (3) above was found to be unstable,
with two tests showing the repeatable results for the threshold speed of instability and a
large sub-synchronous forward whirl instability above the first critical speed. However,
in the second test, the amplitude of instability grew large suddenly and wrecked the
entire test rig. The first critical speed of the rotor was around 5500 rpm.
The experimental results for the tests where the rotor-bearing system became
unstable are described as follows:
31
Fig. 18 Waterfall plot showing threshold speed of instability at 9600 rpm
Fig.18 shows the waterfall plot with the instability threshold at 9600 rpm. Fig.18
shows that the instability grew larger than the synchronous component (the 1X
component is due to imbalance). The plot is a coast-down plot. The instability
disappeared at 9100 rpm. The instability appeared as the loose shrink fit end was heated
for about 8 minutes using the heat guns while the rotor speed was held constant at 9600
rpm.
This test was repeatable under identical conditions. However, in the second test,
the rotor wrecked, as the rotor was coasted-down. The results for this experiment could
be acquired using only the LVTRC data acquisition software, as the ADRE data
acquisition software has some limitations on its file size. The snapshot taken from the
LVTRC screen is shown in Fig.19:
1X
Sub-synchronous component
Sub-synchronous component
32
Fig. 19 Spectrum plot from LVTRC showing the sub-synchronous instability component
Fig.20 shows the picture of the wrecked two-disk rotor. The shaft is completely
bent. The experimental results on the two-disk rotor show that the rotordynamic
instability due to internal friction caused by slipping at the shrink fit interfaces can be
potentially catastrophic.
1XSub-synchronous component
33
Fig. 20 Wrecked two-disk rotor
34
CHAPTER III
INTERNAL FRICTION MOMENTS MODEL
This chapter describes a conceptual model of internal friction developed at the
shrink fit interfaces in rotating machines. This chapter will show that the internal friction
at the shrink fit interface due to relative sliding between the rotating and whirling
mechanical components such as a shaft and a disk, or a disk and a sleeve gives rise to a
system of moments (couples) that are internal to the mechanical system. These moments
are internal because they occur in opposite pairs due to relative sliding between the
rotating (and whirling) mechanical elements as described above. In addition, the internal
moments are generated as a result of the motion of the rotor itself and not otherwise, just
like an imbalance in the rotor exerts dynamic forces on the rotor system when the rotor
executes the rotational motion and does not act on the rotor when the rotor is not
rotating. In other words, the internal friction moments are not applied externally to the
rotating system; instead they are the result of the rotor motion itself which can give rise
to instability or self-excited motion of the rotor-bearing system above the first critical
speed, and can lead to catastrophic failures of the rotor-bearing system, as shown
experimentally in Chapter II.
Before describing the internal friction moments model, a widely known and used
model of internal friction is described to provide some background of the analysis of
internal friction. This model was initially postulated by Kimball [4].It was explained and
expanded in greater analytical detail by Gunter [5]. Although useful and easy to
implement in most rotordynamic computer codes to assess stability of the rotor-bearing
systems with hysteretic and shrink fit friction, the model has a flaw of being physically
inconsistent with the principles of mechanics. On the other hand, the internal moments
model, though not widely used or known, has the virtue of being realistic and consistent
with the principles of mechanics.
35
GUNTER’S FOLLOWER FORCE MODEL
Gunter [5] analyzed an extended Jeffcott rotor model. The word “extended”
means that in the mathematical analysis, the internal friction force acts at the geometric
centre of the disk. Besides internal friction, the rotor foundation and bearings are
assumed to have flexibility and damping properties, in addition to the shaft flexibility.
The extended Jeffcott rotor model is shown in Fig.21:
Fig. 21 Extended Jeffcott rotor model [4]
As discussed in Appendix A of reference [5], Gunter modeled the internal
friction due to shrink fits and other types of friction producing joints, besides the rotor
material hysteresis, as a system of longitudinal stresses similar to the elastic stresses of
the shaft, but instead dependent on the rate of change of strain of the shaft fibers. The
friction forces in case of material hysteresis arise from the dynamic stretching of
36
material elements, whereas in the case of shrink fits, the longitudinal stresses are
developed at the interfaces due to relative sliding between the shrink fit components.
The total longitudinal stresses are assumed as the conventional elastic term which comes
from the beam theory plus a strain rate term. This in effect models the internal hysteresis
of the material as viscous damping. Gunter postulated that the shrink fit internal friction
can be modeled in the same way as the material internal hysteresis, with the magnitude
of shrink fit stresses many times larger than those produced by the material internal
hysteresis. The equivalent moments can be depicted on a cross-section of the rotating
and whirling shaft in Fig. 22 as follows:
Fig. 22 Cross-section of the extended Jeffcott rotor showing the moment and force vectors
Mφ
Location of maximum rate of strain in compression
Location of maximum rate of strain in tension
External damping force vector
Whirl direction
ω
MR
Moment vectors due to rotor’s elastic and hysteresis effects
O X
Y
37
Fig. 22 shows the moment vectors and the external damping force vector acting
on the rotor. The moment vector MR is the result of shaft hysteresis (which will tend to
bend the shaft in the direction of the forward whirl or backward whirl, depending upon
whether the rotational speed is larger or smaller than the whirling speed, respectively)
whereas the moment vector Mφ is the reaction to rotor elastic deformation. The direction
of the moment vector MR in Fig.22 is valid for the case when the rotational speed is
larger than the whirling speed (sub-synchronous whirling). The direction of the moment
vector MR will be reversed if the rotational speed is smaller than the whirling speed.
Gunter postulated that the moment vector MR is equivalent to a follower force
which acts tangential to the whirl orbit and acts as a de-stabilizing or “energy adding”
force when the rotor rotates above the first critical speed, and that the follower force acts
as a stabilizing or damping force when the rotor rotates below its first critical speed.
According to Gunter, the rotor cross-section with an equivalent tangential follower force
looks as shown in Fig.23:
Fig. 23 Free-body diagram of a rotor with internal friction, according to Gunter [4]
X
Y
O
-Kr
C
Fφ
Fd
Whirl orbit
ω
φ
38
As shown in Fig.23, the internal friction follower force Fφ acts in opposition to
the external damping force Fd above the first critical speed of the rotor and tends to drive
the rotor unstable, if the external damping is smaller than the internal friction force.
Below the first critical speed, the follower force reverses its direction and acts as a
damping force. This explains the experimental observations made by Newkirk [2] and
Kimball [3]. However, this model is physically incorrect as explained below:
According to Newton’s Third Law of Motion, for every action there is an equal
and opposite collinear reaction. If a follower force acts on the rotor’s geometric centre as
shown in Fig.23, then according to Newton’s Third Law, an equal and opposite collinear
force must act on a physical attachment or component of the rotor. However, in a
Jeffcott rotor, there is no physical connection from ground to the rotor i.e, to the disk. To
assume that a follower force acts on the rotor due to internal friction or material
hysteresis is the equivalent of assuming the force to be a bearing force reacting to the
ground. In other words, the follower force model can be considered as if there is a
bearing connected through the rotor’s geometric centre to the ground, which is clearly
incorrect, since in an actual physical situation, there is no bearing through the rotor’s
geometric centre to the ground. The only physical connections to the rotor are the
support bearings at the two ends of the rotor. Therefore, the follower force can not
physically exist.
However, Gunter’s model is widely used in industry and research to model the
internal friction. An example can be given of how the internal friction is modeled in
industry by connecting a bearing to the ground through a rotor’s geometric centre at the
shrink fit interface, as shown in Fig.24:
39
Fig. 24 Modeling of internal friction using Gunter’s model
As shown in Fig.24, modeling a rotor system with internal friction using
Gunter’s follower force model requires applying a cross-coupled follower force to the
centre of the disk. To apply this model in XLTRCTM requires applying a user-defined
cross-coupled force at the centre of the disk. This force is applied as a bearing
connecting the rotor disk to the ground. Fig.24 clearly shows the physically incorrect
concept of a “bearing to the ground” to model internal friction.
Therefore, instead of a follower force, it is an internal bending moment, labeled
MR in Fig.22, which acts on the rotor with hysteresis or shrink fit friction. This moment
will tend to bend the rotor in the direction of forward whirl (perpendicular to the
direction of shaft deflection vector) when the rotor operates above the first critical speed.
26 PM25 PM
Shaft124
2016
12
84Shaft1
1
-16
-12
-8
-4
0
4
8
12
16
0 8 16 24 32 40 48
Axial Location, inches
Shaf
t Rad
ius,
inc
hes
Shell Rotor test rig
Modeling the Interference fit
Follower force acting through a bearing to the ground, a physically incorrect model.
40
Below the first critical speed, the internal moment will tend to bend the shaft in the
direction of backward whirl. The internal moments model as described in the next
section addresses the inadequacy of the follower force model.
INTERNAL MOMENTS MODEL
A model to describe the action of internal friction due to the effects of shrink fits
and how it can produce a de-stabilizing “internal moment” is explained by considering
an example of a two-disk rotor as shown below:
Fig. 25 A two-disk rotor whirling in first mode, along with a stiff aluminum sleeve
(vibration of the shaft shown exaggerated to clarify explanation)
Fig.25 shows a two-disk rotor’s solid model with a stiff aluminum sleeve press-
fitted onto the steel disks. The discussion presented is qualitative in nature and
quantitative models are described in Chapter IV. In Fig.25, the rotor is shown to whirl or
vibrate in its first mode (during which the rotor’s centerline assumes a nearly half-
sinusoidal shape) and the exaggerated gap between the faces of the sleeve and the steel
Z
X
Y
ω
Shaft vibrating in the first mode
Stiff aluminum sleeve
Rotor supported at the two ends on bearings
41
disks shows that the disk is slipping at the interface, although total contact is not lost.
Due to relative slipping between the disk and the sleeve, friction forces are generated at
the side positions of the steel disks (plus and minus X locations). The friction forces
occur in equal and opposite pairs (according to Newton’s Third Law of Motion, which
implies that the disks exert equal and opposite forces on the sleeve at the same
locations). The friction forces act in the sense that since one side of the disk is slipping
out, the friction force on it is opposite, whereas for the other side, the friction force will
act so that the combined effect is a couple. An opposite couple acts on the other steel
disk. This can be drawn as follows:
Fig. 26 Free-body diagram of the shaft carrying steel wheels
-Ff
Ff
Ff
-Ff
42
Fig.26 shows that although the net forces due to friction effects cancel out each
other, there is still a bending effect due to these forces, since the forces form a system of
couples.
1. Spin speed larger than whirl speed
Although the couples acting on the two disks are equal in magnitude and
opposite in direction, they tend to bend the shaft in the direction of forward whirl. This
bending of the shaft can maintain the forward whirl and instability of the rotor bearing
system due to internal friction can occur. As the rotor traverses one whirl cycle, the
largest friction forces act at the points of maximum slipping velocity on the disk against
the direction of relative slip. As the disk completes a 90 degree rotation (spin relative to
the whirl vector), the top portion tends to come inside the sleeve (because now it is
moving toward the compression side of the shaft). While tending to come inside the
sleeve, it experiences a friction force that opposes this motion. Similar and vice versa is
the case for a bottom point on the disk. It will experience a friction force as it comes
around towards the top. These forces will create a system of couples that tend to bend
the shaft in the direction of forward whirl. Therefore, it can be seen that the instability
due to internal friction can be represented suitably by means of internal acting moments
that tend to bend the rotor shaft in the direction of forward whirl. These internal
moments can be depicted in the free-body diagram of the two-disk shaft as shown in
Fig.27:
43
Fig. 27 Free-body diagram of the shaft showing equivalent internal friction moment
vectors 2. Spin speed smaller than whirl speed
In this case, friction forces that act on the steel disks act in opposite way to the
one described above. The reason is this that since the whirl speed is now larger than spin
speed, the tension and compression portions (concave and convex sides) of the shaft are
superseding the spin-rotated points on the disks, namely top and bottom points. As the
shaft spins to one-half of the rotation, the top portion will be towards (or approaching)
tension side of the shaft again. This means that it will tend to stick out of the sleeve and
experience a friction force due to loose shrink fit in the direction, as shown in Fig.28.
Similar and vice versa will be the case for a point on the bottom of the disk. Thus a
system of equal and opposite forces will form on the surface of the disk as shown in
Fig.28. This system of forces will create equal and opposite couples on the two disks,
which will be as shown below:
Mf
-Mf
44
Fig. 28 Free-body diagram of the shaft showing equivalent internal friction moment
vectors
Fig.28 shows that the internal friction moments are acting in a way that will tend
to bend the shaft in the direction of backward whirl. Thus, for a spin speed smaller than
the whirl speed, the internal friction moments will damp out the whirling and have a
stabilizing effect.
-Mf
Mf
45
CHAPTER IV
EQUATIONS OF CROSS-COUPLED MOMENTS FOR THREE INTERFACE
FRICTION MODELS
The mathematical analysis for deriving the expressions for cross-coupled forces
or moments for various internal friction models was carried out by some researchers,
such as Gunter [4], Walton [5, 6], Lund [8] and Black [10]. In these references, the
interface friction models which were primarily considered were viscous friction,
Coulomb friction and hysteretic friction models and their effects on rotordynamic
stability were analyzed. In addition, Lund analyzed a micro-slip model, which is mainly
applicable for an analysis of slip in axial spline joints. Both Black and Gunter formulated
the de-stabilizing mechanism as a follower force, which is a physically incorrect model
but still provides some useful insight into the nature of rotordynamic stability due to
internal friction. However, with the availability of high-speed computers and
comprehensive rotordynamic analysis softwares such as XLTRCTM, it is now possible to
analyze a physically correct model of the problem by including the internal cross-
coupled moments at the interface, rather than equivalent external follower forces, a
procedure that has been followed widely in recent times. Even though the internal cross-
coupled moments model can be implemented using the XLTRCTM software, the software
has a limitation of accepting only linear models for the forces and the moments.
46
The analysis for equations of cross-coupled moments for various interface
friction models (except for the Coulomb friction model) which is developed in this
chapter is drawn mainly from Lund’s and Walton’s work, with figures to help illustrate
the derivation of cross-coupled moment equations. In addition, this chapter will explain
each of these models in some detail and highlight the significance of models from point
of view of how realistic their predictions are and how practical it is to implement them
using the XLTRCTM software. It will be shown that of the three models, only one model,
namely the viscous friction, is the most practical in terms of its ease of implementation
in the XLTRCTM software, because it is a linear model. However, a drawback of using
the viscous friction model is that although it will predict the threshold speeds of
instability and give an overall knowledge of the stability of the rotor-bearing system, the
predictions are not completely accurate, in that they predict an unlimited range of
instability above the threshold speed of instability, which is in contradiction with the
experimental research on internal friction [14]. On the other hand, even though
Coulomb friction and the hysteretic friction models have the virtue of being realistic, as
described in the Literature Review using the references [10] and [5,6,14], it is
impractical to implement them in the XLTRCTM software due to their non-linear
character.
47
BASIC ROTORDYNAMIC MODEL FOR ANALYSIS
Fig. 29 Sketch of a flexible vibrating shaft with a shrink fitted sleeve. The geometry and kinematics are shown.
To derive the equations for cross-coupled moments that are developed due to
internal friction at the shrink fit interfaces, consider a rotor model as shown in Fig.29.
The model shown is useful in developing the equations of cross-coupled moments for
the viscous friction, the Coulomb friction and the hysteretic friction models.
X
Y
Z
ω
x y
z
A sleeve shrink fitted to the shaft
A flexible vibrating shaft with a shrink fit sleeve
Rotor supported at the two ends on bearings
θ
φ
A
B
48
In Fig.29, ‘OXYZ’ is an inertial or a fixed frame of reference, whereas ‘oxyz’ is
a rotating frame of reference which is attached to the rotor and which is rotating at a
speed ‘ω’ with respect to the fixed frame of reference.
The basic kinematics of a joint such as a shrink fitted joint in a rotor is
characterized by a discontinuity in slope of the rotor at the juncture of the joint. In
Fig.29, the discontinuity in slope is shown where the shaft interfaces with the sleeve,
such as at interfaces A and B. This is illustrated more clearly in the side view of the rotor
as shown in Fig.30 below:
Fig. 30 Side view of the rotor model from Fig.29, showing the discontinuity of slope at
the shrink fit interface between the shaft and the sleeve.
The change in slope at the shrink fit interface occurs because the shrink fitted
sleeve may have different material and geometric properties as compared to the shaft due
to which it may be stiffer in bending as compared to the shaft on which it is mounted
x
z
Y
Z X
θ
ω
49
through the shrink fit. Therefore, at the shrink fit interface, the sleeve will not allow the
shaft to bend as much as it would if there were no sleeve mounted on it. This will result
in a difference in slope between the “interface free” and the “inside the interface”
segments of the shaft. In Fig.29 and Fig.30, this difference in slopes is shown as angular
displacements ‘θ’ and ‘φ’ about the X and the Y axes, respectively.
It follows that if there was a micro-slip between the shaft and the sleeve at the
interface, it can be quantified using the angular displacements coordinates ‘θ’ and ‘φ’.
Since the micro-slip motion is described using the angular coordinates about the fixed
OXYZ coordinate system, it follows that associated with this micro-slip angular motion
at the interface will be corresponding friction bending moments, or couples, which will
be developed due to the micro-slip angular motion of the shaft at the shrink fit interface.
The couple is developed due to friction forces acting on the periphery of the shaft, which
occur in equal and opposite pairs. That is, the diametrically opposite directions on the
periphery of the disk and the shaft have friction force pairs equal in magnitude and
opposite in direction due to slipping motion of the shaft that will be equivalent to a
couple acting on the shaft, tending to bend it either in the direction of forward whirl (at
supercritical speeds) or backward whirl (at sub critical speeds).
The generation of moments at the interface due to this angular micro-slip can be
shown schematically as if there were a torsional spring and a damper at the interface
between the shaft and the sleeve. The presence of torsional elements gives rise to the
development of cross-coupled moments at the interface due to the micro-slip. This is
shown in Fig. 31 on the next page:
50
Fig. 31 A model of the shrink fit interface friction showing the spring elements. The
torsional springs account for the cross-coupled moments.
In Fig.31, only torsional springs (but not the dampers) are shown for clarity,
although the presence of dampers is implied. In addition, the translational springs in
orthogonal directions are shown at the interface. The presence of translational springs
implies the existence of a reasonably tight fit (with correspondingly high values of
stiffness coefficients) that will not allow any substantial relative translational motion at
the interface. Fig.31 is the model on which the XLTRC simulations for a rotor-bearing
stability are based in this dissertation work.
In order to develop the equations for cross-coupled moments, consider the
transformation for slopes ‘θ’ and ‘φ’ from fixed to the rotating frame of reference.
Measured from the rotating frame of reference, if the differences in slopes at the
interface are ‘α’ and ‘β’ about the x and the y axes respectively, the transformation can
be derived using Fig.32 as shown below:
A pitch-yaw spring element
A translational spring element
X
Y
Z
51
Fig. 32 Coordinate transformation between the fixed and the rotating frames of reference
)tsin()tcos( ωβωαθ −= (3)
)tcos()tsin( ωβωαφ += (4)
In equations (3) and (4), ‘ω’ is the angular speed of rotation and ‘t’ is time.
From equations (3) and (4), the following equations can be written:
)tsin()tcos( ωβωαωφθ•••
−=+ (5)
)tcos()tsin( ωβωαωθφ•••
+=− (6)
The same transformation applies to the bending moment components (which
arise due to micro-slip motion) measured in the fixed and the rotating frames of
reference. The transformation from the fixed to the rotating frame can be expressed as
follows:
X
Y
x y
ωt
θ
φ
α β
52
)tsin(M)tcos(MM ωω βαθ −= (7)
)tcos(M)tsin(MM ωω βαφ += (8)
KINEMATICS OF ROTOR MOTION
The motion is assumed to be harmonic with angular frequency ‘Ω’ (precession
speed) such that:
e)iRe()tsin()tcos()t( tiSCSC
ΩθθΩθΩθθ +=−= (9)
In equation (9), the precession motion of rotor ‘θ(t)’ (angular motion about the
X-axis) is defined in terms of the precession frequency ‘Ω’. The physical interpretation
of ‘Ω’ is the whirling frequency of the rotor, as would be measured through a signal
analyzer when sub-synchronous vibrations are excited in the rotor. In equation (9),
complex variables are used as an alternate and a convenient method to express the
motion variables in a compact form. The symbol ‘i’ is for the complex variable operator:
1i −=
In accordance with the usual convention θ(t) can be expressed in a compact form
as follows:
SC iθθθ += (10)
In writing equation (10), it is assumed that the real operator ‘Re’ and complex
exponential function ‘eiΩt’ are implicit and they will be assumed to always apply, even
though not shown in the expression, whenever a kinematic variable is expressed in its
complex form.
53
The angular motion due to the micro-slip friction moments of the rotor, θ(t), can
be expressed in a more general form as follows:
bf θθθ += (11)
In equation (11), ‘θf’ and ‘θb’ are complex numbers. The subscripts denote
“forward” and “backward” components, respectively. Equation (11) expresses the
concept that the motion may be thought of as made up of two circular whirl motions,
once with the forward whirl, ‘θf’, and another one with the backward whirl, ‘θb’.
Without loss of generality, the micro-slip angular motion, φ(t), of the rotor about
the Y-axis can be expressed as follow:
bf ii θθφ +−= (12)
The same convention applies to φ(t) in equation (12) as is applied to θ(t) in
equation (11) (that the real operator ‘Re’ and the complex exponential function ‘eiΩt’ are
assumed implicit and are not shown exclusively in writing equation (12)).
From equations (11) and (12), the forward and backward components, θf and θb
can be expressed as follows:
)i(21
f φθθ += (13)
)i(21
b φθθ −= (14)
Using the coordinate transformations (equations (3) and (4)) in conjunction with
equations (11) and (12), the following equations result:
t)(i
bt)(i
f ee ωΩωΩ θθα +− += (15)
54
t)(i
bt)(i
f eiei ωΩωΩ θθβ +− +−= (16)
Equations (15) and (16) show that the motion which in the fixed frame has only
the single frequency Ω, but with an elliptical orbit, splits up into two circular orbits, each
with its own frequency, in the rotating frame.
1. Viscous friction model
Consider the moment stiffness of the shrink fit joint to be ‘K’ and damping at the
interface to be ‘C’. If the internal friction is modeled as viscous friction, the bending
moment components (due to micro-slip) as measured in the rotating frame of reference
‘oxyz’ can be expressed as:
•
+= ααα CKM (17)
•+= βββ CKM (18)
By using the coordinate transformations from equations (3),(4),(5),(6),(7) and
(8), the components of bending moments due to micro-slip in the fixed reference frame
‘OXYZ’ can be expressed as follows:
)(CKM ωφθθθ ++=•
(19)
)(CKM ωθφφφ −+=•
(20)
55
From equations (19) and (20), the bending moment components equations show
the presence of cross-coupling stiffness terms ‘Cω’, which being of opposite sign, act
destabilizing [8]. This can be shown by computing the energy dissipated in one cycle:
])()[(C2dt)MM(E 2b
2f
/2
0diss θωΩθωΩπφθ
ωπ
φθ ++−=+= ∫••
(21)
In deriving equation (21), the angular velocities of the micro-slip rotor motion
about the X and the Y axes are calculated using equations (3), (4), (15) and (16).
Equation (21) shows that when the rotational speed ‘ω’ exceeds the natural
frequency ‘Ω’, the first term becomes negative. Therefore, if the whirl mode is a circular
orbit with forward whirl (θb = 0), the rotor becomes unstable when ω = Ω. If the
backward whirl component is not equal to zero, then the instability will occur depending
upon the speed, when the first term exceeds the second term in equation (21).
56
2. Coulomb friction model
Fig. 33 End view of the disk showing the radial stresses acting on its surface
To formulate the internal friction moments using the Coulomb friction model for
a rotor disk which is spinning, whirling and nutating (tilting), consider the plane of the
disk and the rotating oxyz frame attached to the disk as shown in Fig.33. Consider
another rotating frame which has its origin at the geometric center ‘C’ of the disk. The
variable ‘ψ’ locates the position of a point on the disk, whereas ‘σr’ denotes the radial
stress acting at a location ‘ψ’ of the disk due to the shrink fit and spin of the shaft.
Differential radial force acting on the disk in its first quadrant (as seen from
o’x’y’z’ frame) can be expressed as follows:
dAdF rr σ= (22)
X
Y
ωt
ψdψ
57
The differential area element over which the radial stress acts can be expressed as
follows:
ψψ RLdL)Rd(dA == (23)
In equation (23), ‘L’ is the axial span of the disk and ‘R’ is its radius. Using
equation (23), equation (22) can be expressed as:
ψσ dRLdF rr = (24)
A corresponding differential friction force acting over the same differential
element of area is developed, based on Coulomb model, such that:
)Vsgn(dRLdF Slidingrz ψσμ= (25)
In equation (25), ‘μ’ is either the static friction coefficient, if VSliding = 0, or it is
the kinetic coefficient of friction, if VSliding ≠ 0. The differential friction force as
expressed in equation (25) acts in the axial direction, because according to Coulomb
friction law, the friction force acts perpendicular to the normal force, which is in the
radial direction (due to the normal radial stress at the interface).
Based on above equations, the differential moment due to normal stress and the
corresponding frictional shear stress on the surface above axis ‘ox’ can be expressed as:
zFdRMd→→→
×= (26)
In equation (26), the vector ‘R’ is moment arm of the differential force dFz, and
its magnitude is equal to the interface radius of the disk. The direction of vector ‘R’ is
radial, with this direction being from geometric center ‘C’ to the periphery of the disk.
Thus, ‘R’ can be expressed as:
58
−
→= reRR (27)
Carrying out the cross product in equation (27), the differential moment vector is
expressed as:
)Vsgn(deLR)Vsgn(eRdFMd Slidingr2
Slidingz ψσμ ψψ−−
→−=−= (28)
In equation (28), the unit vector in ‘ψ’ direction is a result of cross product of the
radial unit vector and the ‘k’ vector, which is perpendicular to the plane of the disk.
From simple geometry, a well-known transformation between the unit vectors in the disk
(that is, er and eψ) and the unit vectors in the rotating oxyz frame, is given by:
ψψ sinjcosier−−−
+= (29)
ψψψ cosjsinie−−−
+−= (30)
Using equations (29) and (30), the differential moment vector can be integrated
from 0 to 2π to give the value of the resultant moment vector due to the differential
frictional force contributions.
To compute the integral of equation (28) in the interval 0 to 2π, it is necessary to
establish an expression for the sliding velocity ‘VSliding’, because the numerical sign of
the differential moment in equation (20) depends upon the sign of the relative sliding
velocity, which will vary from positive to negative along the circumference of the disk
(the relative sliding velocity will switch signs from +1 to -1 in the interval 0 to 2π).
Relative sliding velocity
To derive an expression for the relative sliding velocity, consider Figs.1 and 5. In
terms of position vectors and angular velocity vectors, the relative sliding velocity of a
point P on the circumference of the disk (or the shaft) can be expressed as follows:
59
−−
→×= PP rV ω (31)
Equation (31) expresses the velocity of the point ‘P’ as measured from point ‘C’
due to angular velocity of the shaft relative to the sleeve due to slip at the interface.
The expression for the angular velocity to be used in equation (31) can be
expressed as (using rotating frame coordinates):
−
•
−
•
−+= ji βαω (32)
The position vector ‘rP’ using the rotating frame ‘oxyz’ can be expressed as
follows:
−−−+= jsinRicosRrP ψψ (33)
Carrying out the cross-product in equation (31) using equations (32) and (33), the
expression for relative sliding velocity of a point ‘P’ on the shaft is:
−
••→−= k)sinRcosR(VP ψβψα (34)
Equation (34) shows that the relative sliding velocity of a point ‘P’ on the disk
will vary sinusoidally as a function of the circumferential location on the disk as the
shaft whirls and as the disk on the shaft undergoes slipping at the shrink fit interface.
The relative sliding velocity in equation (34) is in the axial direction (in the
direction of the unit vector ‘k’).
From equation (34), the magnitude of relative sliding velocity can be expressed
as follows:
60
)sin(RsinRcosRV22
Sliding γψβαψβψα −+=−=••••
(35)
In equation (35), the argument ‘γ’ is defined as follows:
22cos
••
•
+
=
βα
αγ (36)
22sin
••
•
+
=
βα
βγ (37)
From equation (35), the sliding velocity changes sign as the sinusoidal function
changes sign. Using equation (35), the following conditions for sign function of the
sliding velocity are defined:
1)Vsgn( Sliding += , γπψγ +<< (38)
1)Vsgn( Sliding −= , γπψγπ +<<+ 2 (39)
Integrating equation (28) in the interval ψ = 0 to 2π by using equations (38) and
(39):
)sinjcosi(LR4MdMdMdM r2
22
0
γγσμγπ
γ
γπ
γπ
π
−−
+ +
+
→→→→+−=+== ∫ ∫∫ (40)
Equation (40) shows that the resultant moment due to frictional stresses and
sliding velocity at the interface is a non-zero vector, with components along both the x
and the y axes of the rotating frame of reference. Equation (40) shows the presence of a
couple that is developed due to frictional forces acting at the interface. The frictional
forces along the circumference reverse signs due to reversal of sign of the sliding
61
velocity (equation (35)) and thus form equal and opposite pairs of forces, which form a
resultant couple as expressed by equation (40).
From equation (40) and equations (7) and (8), the frictional moment has
components along both the X and the Y axes of the fixed frame of reference.
From equation (40), the components of frictional moments along the x and the y
axes can be defined as follows:
22r
2 LR4M••
•
+
−=
βα
ασμα (42)
22r
2 LR4M••
•
+
−=
βα
βσμβ (43)
Using equations (5),(6), (7) and (8) and applying them to equations (42) and (43)
to obtain the components of the bending moments in the X and the Y directions, the
following equations are obtained:
22r2 )(LR4M
••
•
+
+−=
βα
ωφθσμθ (44)
22r
2 )(LR4M••
•
+
−−=
βα
ωθφσμφ (45)
Equations (44) and (45) can be expressed completely in terms of the fixed frame
coordinates as follows (using equations (3) and (4)):
62
22r
2
)()(
)(LR4M
ωθφωφθ
ωφθσμθ
−++
+−=
••
•
(46)
22r
2
)()(
)(LR4M
ωθφωφθ
ωθφσμφ
−++
−−=
••
•
(47)
Equations (46) and (47) show that the moment components due to Coulomb
friction at the shrink fit interface are cross-coupled moments, due to the presence of
speed dependent terms in the numerators, which are of opposite signs in the two
equations. The internal moments in Coulomb friction model are non-linear functions of
the rotating speed and amplitudes as well as angular velocities of the micro-slip. This is
in contrast with the viscous friction model (equations (19) and (20)) where the cross-
coupled moments are linear functions of the rotational speed as well as the amplitude
and angular velocity of the micro-slip.
Therefore, the formulation shows that due to reversal of the friction force
direction over the periphery of the disk, the resultant moment is not zero. The resultant
moment depends upon several geometric and dynamic parameters of the rotor, such as
the radius, axial span, friction coefficient, the normal radial stress (which in turn is a
function of geometric and elastic properties of the disk, as well as the value of shrink fit
at zero speed and the spin speed) and sign of the sliding velocity.
3. Hysteretic friction model
The hysteretic friction or the solid friction model assumes the interface internal
friction moments to be of the following mathematical form:
αααα )sgn(CKM•
+= (48)
ββββ )sgn(CKM•
+= (49)
63
In equations (48) and (49), ‘K’ is the moment stiffness coefficient whereas ‘C’ is
also a moment stiffness coefficient, but it differs from ‘K’ in the sense that it contributes
alternately positively and negatively to the moments in each direction, depending upon
the frequency of micro-slip motion.
Using equations (7),(8), (15) and (16), the corresponding components of bending
moments developed at the interface about the X and the Y axes can be expressed as
follows:
φωΩωΩθωΩωΩθ ))sgn()(sgn(C21))]sgn()(sgn(C
21iK[M −−++−+++= (50)
θωΩωΩφωΩωΩφ ))sgn()(sgn(C21))]sgn()(sgn(C
21iK[M −−+−−+++= (51)
From equations (50) and (51), there are cross-coupling terms in the expressions
for moment components. These cross-coupled terms produce destabilizing motion of the
rotor. As calculated for the case of viscous friction model in equation (13), the energy
dissipated per cycle can also be calculated for the case of hysteretic friction model. It is
expressed by the following equation:
])sgn()[sgn(C2E 2b
2fdiss θωΩθωΩπ ++−= (52)
Equation (52) shows that the energy dissipated per cycle depends upon the sign
of the term that involves the frequency difference between whirling frequency and the
rotational speed. When rotational speed exceeds the whirling frequency, the energy is
“added” to the system, instead of being dissipated, provided the backward whirl
component is smaller in magnitude as compared to the forward whirl component.
However, the amount of energy added to the rotor or dissipated from the rotor due to
slippage at the interface is independent of magnitude of whirling frequency. The
independence of energy dissipation or addition from the magnitude of whirling
64
frequency is in better agreement with experiments, as compared to the viscous friction
model prediction, in which case energy dissipation or addition is dependent upon the
magnitude of whirling frequency, in addition to the numerical sign of frequency
difference term Ω-ω.
PHYSICAL INTERPRETATION OF FRICTION MODELS
The various friction models discussed in previous pages can be interpreted
physically based on the concept of energy dissipation per cycle for each of the models.
The qualitative and quantitative description of each of the three models is as follows:
1. Viscous friction model
Viscous friction model is one of the most commonly used models in vibrations
and rotordynamic analysis. The viscous friction model is especially useful because of its
linear mathematical form. The governing differential equations for most mechanical
systems with viscous damping terms have exact analytical solutions.
When the viscous friction model is applied to study the phenomenon of internal
friction in solids, the governing physical model of the solid is called as Kelvin-Voigt
model. In schematic form, such a model for a solid with internal friction is shown in
Fig.34:
65
Fig. 34 Kelvin-Voigt model of internal friction in solids
This model is also commonly termed as viscoelastic model. From Fig.34, the
model takes into account both elastic (energy absorbing and recovering) as well as
friction (energy dissipation) behavior of a solid subjected to stresses. The constitutive
equation of Kelvin model can be written as follows:
•
+= εμεσ E (53)
In equation (53), the applied stress on the body is σ, whereas ε is the strain
induced in the body. The second term on the right hand side of equation (53) is the
energy dissipation term due to strain rate. The elastic modulus is denoted by E.
To derive an expression for energy dissipated per cycle, consider the solid
subjected to harmonically varying strain:
)tsin()t( 0 ωεε = (54)
Differentiating equation (52) with respect to time t, the equation for strain rate is:
μ
σ σ
E
66
2/1220
2/1200 )()]t(sin1[)tcos( εεωωεωωεε −±=−±==
• (55)
Substituting equation (55) in equation (53) leads to the following equation:
2/122
0 )(E εεμωεσ −±= (56)
The graph of equation (56) is an ellipse, with stress σ on vertical axis and the
strain ε on horizontal axis. The area of the ellipse is energy dissipated per cycle. The
graph of equation (56), which is called as hysteresis loop, is shown on the next page:
Fig. 35 Hysteresis loop due to viscous friction
ε
σ
ε
ε0
-ε0
dε/dt >0
dε/dt <0
σ = Eε
σ
67
Fig.35 shows the hysteresis loop formed by plotting equation (56). The major
axis of the ellipse is the line: σ = Eε. The upper part of the ellipse corresponds to 0>•ε ,
because εσ E> in that case. The lower part corresponds to 0<•ε .
The energy dissipated per unit area per cycle can be obtained by calculating the
area of the hysteresis loop. The energy dissipated per area per cycle is:
20
2/2
0diss dtE πωμεεμ
ωπ
==•
∫ (57)
From equation (57), the energy dissipated is proportional to the frequency of
oscillation ω. This dependence on frequency comes about as a result of the term•εμ ,
which is proportional to ω.
The same concept of viscous friction can be extended to conceptualize interface
friction moments in terms of a direct stiffness coefficient K and a direct damping
coefficient C. The energy dissipated or added to the rotor system in the case of viscous
friction model is proportional to the frequency difference, as shown in equation (13).
Therefore, the viscous friction model assumes that the shrink fit interface friction
moments are comprised of a direct spring effect, whereby the spring tends to restore the
relative slip motion of the shaft, whereas simultaneously, there is an energy dissipation
effect, the energy of micro-slip motion is dissipated or added, depending upon the
difference of whirling frequency and the rotational speed. This can be depicted as shown
in Fig.36:
68
Fig. 36 Schematic description of a shrink fit interface using a torsional spring and a
damper
2. Hysteretic friction model
Hysteretic friction model is less widely utilized in modeling damping and friction
in mechanical systems, as compared to the viscous friction model. This is because it is a
non-linear model. Nevertheless, the model is important, because the rotordynamic
predictions of hysteretic friction model for slippage in the shrink fits joints are validated
by experiments [14]. To explain a physical model for hysteretic friction, the following
development is adopted from [23].
In schematic form, a hysteretic friction model can be represented as the following
mechanical model:
K
C
αX
69
Fig. 37 A mechanical model for illustration of the hysteretic friction
As shown in Fig.37, consider a spring-mass system which is subjected to
horizontal forces F. In addition, the mass is loaded by a vertical force N. The mass rests
on a surface that has friction, which opposes the displacement of mass relative to the
surface, when the horizontal forces are applied. The friction can be Coulomb friction or
any other model which models the resistance of mass to sliding relative to the surface.
Initially, when the forces F are applied to the system, the spring of stiffness K
will stretch or compress (if forces are in opposite direction to that shown in Fig.37), and
the mass will not move, owing to the resistance offered by the surface on which it rests.
The deflection in the spring is F/K only. As the force is gradually increased, a limiting
value of the force (Fmax) will reach at which the mass eventually starts to slide on the
surface. In this case, the deflection of the spring will be equal to the deflection of the
mass plus the static deflection F/K, with the same amount of limiting amount of applied
force Fmax. This is the case of increase or decrease of deformation of the spring without
accompanying increase in the magnitude of limiting amount of force. In addition, the
spring deformation (stretch or compression) depends on the sign of the relative velocity
of the mass and the surface, and not on the magnitude of relative velocity. Therefore, the
model in Fig.37 is a non-linear model.
K
F F
m
N X
Y
70
As in the case of viscous friction model, a hysteresis loop can be drawn for such
a model. Replacing the forces by stresses and the spring deformations by strains, the
following figure illustrates the variation of applied stresses with accompanying strains in
the system:
Fig. 38 Hysteresis loop for hysteretic friction model
Fig.38 shows the hysteresis loop for the mechanical model in Fig.37. From point
‘0’ to ‘1’, the only strain of the spring is due to the elastic strain, σ/E. As the limiting
value of the stress σmax is reached, the mass starts sliding on the surface and the strain in
the spring increases as a result of the sliding, with no accompanying increase of stress.
This is indicated in Fig.38 as part of the graph from point ‘1’ to ‘2’. After a certain
maximum strain ε0 is reached, the spring force starts backward motion of the mass,
which results in the decrease of strain, and is shown by line segment ‘2’ to ‘3’ in Fig.38.
ε
σmax
-σmax
ε0
-ε0
0
1 2
34
5
σ
71
Between points ‘2’ to ‘3’, the velocity of the mass is in the negative X direction. The
path ‘2’ to ‘3’ is a straight line, because the spring strain is a linear function of stress. At
point ‘3’, a maximum stress -σmax is reached. It can be seen that the hysteresis loop
formed due to hysteretic friction is dependent upon the sign or sense of the velocity, but
not on its magnitude.
As the mass is acted upon by the negative maximum stress -σmax, it slides and the
spring compresses at the same strain due to sliding motion. This is shown as points ‘3’
and ‘4’. When the spring undergoes a maximum compressive strain of -ε0, the spring
force tends to move the mass in the positive X direction. This is shown as segment ‘4’ to
‘5’. In this part of the loop, the velocity of the mass is in the positive X direction.
Finally, the maximum stress σmax is reached at point ‘5’ and from ‘5’ to ‘1’, the mass
undergoes sliding and the cycle continues.
For hysteretic friction, it is concluded that the amount of energy dissipated is
dependent on the magnitude of strain, but not on the magnitude of strain rate. It is,
however, dependent upon the sense of strain rate (positive or negative).
3. Coulomb friction model
Coulomb friction model is one of the most widely known, but lesser widely used
model (except for analysis of relatively simple dynamics problems) for analysis of
friction between dry surfaces. It is known for its better agreement with experiments of
relative motion between solid objects as compared to any other model, but it is less used
due to its non-linear character. As shown in [14], the experimental test rotor system with
shrink fit interface joints exhibits a damping character that can be explained with
Coulomb and hysteretic damping models, but less well by the viscous damping model.
To explain the physical model of Coulomb damping, the following development
is extensively adopted from [24].
Consider a solid block of weight W resting on a horizontal surface as shown in
Fig.39 below:
72
Fig. 39 Mechanical model to explain Coulomb friction
As shown in Fig.39, the block is subjected to a horizontal force P that tends to
pull the block in the direction of application of force. In the model considered, the
magnitude of P varies continuously from zero to a value sufficient to move the block and
give it an appreciable velocity. The free-body diagram of the block for any value of P is
also shown in Fig.39. In the free-body diagram, the tangential friction force exerted by
the plane on the block is labeled F. This friction force will always be in a direction to
oppose motion or the tendency toward motion of the block. There is also a normal force
N which in this case is equal to W, and the total force R exerted by the supporting
surface on the block is the resultant of N and F.
A magnified view of the irregularities of the mating surfaces will aid in
visualizing the mechanical action of friction. The magnified view is shown in Fig.40 on
the next page:
P W
F
N α R
W P
73
Fig. 40 A magnified view of the irregularities at the mating surfaces
Fig.40 shows that the support is necessarily intermittent and exists at the mating
humps. The direction of each of the reactions on the block R1, R2, R3 etc., will depend
not only on the geometric profile of the irregularities but also on the extent of local
deformations, as well as the welding that can take place on a minute scale at each contact
point. The total normal force N is the sum of the n-components of the R’s, and the total
frictional force F is the sum of the t-components of the R’s. When the surfaces are in
relative motion, the contacts are more nearly along the tops of the humps, and the t-
components of the R’s will be smaller than when the surfaces are at rest relative to one
another. This consideration helps to explain the known fact that the force P necessary to
maintain motion is less than that required to start the block when the irregularities are
more nearly in mesh.
In the model of Fig.39, assume that the friction force F is measured as a function
of P. The resulting experimental relation is indicated in Fig.41 on the next page:
t
n
R3 R1 R2
74
Fig. 41 Friction force F as a function of applied force P
From Fig.41, when P is zero, equilibrium requires that there be no friction force.
As P is increased, the friction force must be equal and opposite to P as long as the block
does not slip. During this period the block is in equilibrium, and all force acting on the
block must satisfy the equilibrium conditions (zero net force and moment). Finally a
value of P is reached which causes the block to slip and to move in the direction of the
applied force. Simultaneously the friction force drops slightly and rather abruptly to a
somewhat lower value. It remains essentially constant for a period but then drops off still
more with higher velocities.
The region up to the point of slippage or impending motion is known as the range
of static friction. This force may have any value from zero up to and including, in the
limit, the maximum value. The magnitude of the maximum static friction force is
determined from the Coulomb law of static friction as follows:
NF SmaxS μ= (58)
F
P
FS = P
Fsmax=μSN Fk=μkN
Static friction (no motion)
Kinetic friction (motion)
75
In equation (58), μS is a constant, called as the coefficient of static friction, which
depends the materials of the mating surfaces and their geometry.
After slippage occurs a condition of kinetic friction is involved. Kinetic friction
force is usually slightly lower than the maximum static friction force in magnitude.
Moreover, the sense or direction of the kinetic friction force depends upon the direction
of relative velocity of the mating surfaces. Coulomb’s law of kinetic friction expresses
the magnitude and sense of the kinetic friction force as follows:
)Vsgn(NF lReKK μ= (59)
Equation (59) is a non-linear equation in the motion variable of the block because
it involves the “signum” function, which is a non-linear function.
76
CHAPTER V
EXPLANATION OF KIMBALL’S EXPERIMENTS USING INTERNAL
MOMENTS MODEL
BASIC THEORY OF ROTOR INTERNAL FRICTION
The fundamental theory to explain the rotor internal friction comes from the
work of A.L. Kimball (1924). For this, consider the following figure:
Fig. 42 A rotor disk supported on a flexible shaft rotating clockwise.
In Fig.42, the rotor disk is a heavy weight W and is supported on a flexible shaft
with internal hysteresis. The reaction loads are acting on the two ends through the
bearings. For simplicity, imbalance is neglected and the downwards deflection of the
flexible shaft is due to gravity alone. If the shaft were purely elastic, the rotor will
deflect vertically downwards. However, when the internal friction is present in the shaft
fibers, the rotor does not deflect vertically downwards, and instead makes an inclination
W W/2
W/2
ω
77
angle φ with the vertical when the shaft spins, as shown in the end-view of the disk in
Fig.43:
Fig. 43 Rotor disk side views (a) Purely elastic deflection (b) Deflection with internal friction
Fig.43 shows that the disk will be deflected sideways. The deflection will be in
the direction of rotation. This can be explained as follows: When the rotor is being
turned at a constant speed, there is applied torque acting on it, in addition to the reactions
from the bearings. Due to bearings, the torque is dissipated, so that the rotor turns at a
constant speed. If the rotor is turned at a speed with the supply torque cut-off, for
example in a rotor coast-down, then still there will be a forward deflection of the shaft
due to rate of change of strain in shaft fibers, which comes about as a result of shaft
W
W
Φ = 0
ω
(a)
W
W
Φ
(b)
ω
78
rotation and the shaft fibers being in tension and compression due to initial sag of the
shaft. As long is there is rotation of the shaft, the material internal hysteresis will be
active. As there are frictional tensile and compressive forces acting on the shaft, they
tend to bend the shaft in the direction of forward whirl, in the same way as an elastic
reaction from the shaft tends to straighten the shaft. Therefore, it is seen that the friction
can drive whirl instability of the system. Since the moment vector is parallel to the plane
of deflection, it follows that it can be replaced by equal and opposite forces
perpendicular to the plane of disk. As a result of these compressive and tensile frictional
forces, the shaft deflects like in Fig. 43(b). This was the fundamental hypothesis of
Kimball to explain forward whirl instability due to internal friction.
MODIFICATION TO KIMBALL’S HYPOTHESIS
As shown in Fig.43 (b), and as explained by Kimball, the internal friction forces
due to material hysteresis are perpendicular to the plane of the disk. If this is so, then the
consideration of the bending of the shaft due to these friction forces, which are
equivalent to a couple moment in the plane of the disk (acting in the vertical direction
upwards), should show bending of the shaft in a way which is not shown in Fig.43 (b),
and it was actually not proposed, shown or discussed by Kimball in his papers. In Fig.43
(b), as the shaft deflects in the direction of forward whirl, at the same time, internal
friction moments will cause it to bend in the following way:
79
Fig. 44 Side view of the disk-shaft under the influence of internal hysteresis
A close-up view of the shaft with frictional tensile and compressive stresses, as
seen from the top (top view) can be depicted as shown in Fig.45 on next page:
W
-W
Bent shaft due to friction moments
ω
Internal friction moment vector
80
Fig. 45 Frictional tensile and compressive stresses acting on the shaft
Fig.45 shows the frictional tensile and compressive stresses acting on elements
inside the shaft, due to material internal hysteresis. The case is for a supercritical speed.
It can be seen that the compressive and the tensile stresses form equivalent bending
moments (frictional moments) to be developed inside the shaft, that tend to bend the
shaft in the direction of forward whirl. Therefore, in addition to the initial elastic sag,
this bending of the shaft due to the internal hysteresis will be present to cause a
deflection, which will look somewhat similar to that as illustrated in Fig.44.
Fig.46 shows the bent centerline of the shaft due to internal friction moments. It
is possible that the measurements made by Kimball, in which he measured the sideways
deflection of a thin, overhung shaft loaded vertically, were actually the measurements of
CL dε/dt >0
dε/dt <0
Shaft deflection caused by internal friction moments
Shaft element in tension
Shaft element in compression
ω
81
this “bent” deflection (as shown in Figs.44 and 45), rather than the postulated sideways
deflection (as shown in Fig.43 (b)).
Fig. 46 Top-view of the rotor showing the bent centerline due to friction moments
In his experiments, Kimball also constructed a single-disk vertical rotor and
observed violent whirling of the rotor (above its first critical speed) when the shaft was
wrapped around with steel mesh wires. The rubbing of wire with the shaft produced the
internal friction. However, he did not perform a similar experiment to assess the strength
of internal rotor hysteresis on rotor stability (no shrinkage fit). Also, his measurements
indicate that when the experimental rotors were shrink fitted with rings, the amount of
sideways deflection was atleast 3 to 4 times stronger than with no shrinkage and that the
shrinkages were the major source of internal friction instability as compared to the rotor
internal hystersis.
From these facts, it can be argued that in shrink fit joints, there are internal
friction forces produced that act perpendicular to the plane of the disk of the rotor. As a
result, these forces form a de-stabilizing couple moment vector in the plane of the disk
CL
Shaft’s bent centerline Internal friction moments bending the shaft
Mf -Mf
ω
82
that tends to bend the shaft in a way similar to that in Figs. 44, 45 and 46. This bending
of the shaft produces the forward whirl sub-synchronous instability of the rotor. The
material internal hysteresis effect of the experimental rotors, as investigated by Kimball,
can therefore be said to lead to the same physical shape of the rotor as shown in Figs.44,
45 and 45.
Therefore, the internal friction moment model can explain Kimball’s
measurements results if it is assumed that his sideways deflection measurements were
actually the measurements of the bent sideways deflection similar to Figs. 44 ,45 and 46,
rather than that in Fig.43 (b). This can be particularly true, because Kimball’s
measurements were made while neither looking at the rotor’s side-view directly, for
example, from some optical instrument, nor the deflections measured could be said to be
totally occurring for a shaft which is not bent from moments other than the elastic
bending moments. Thus, the proposed modification to Kimball’s hypothesis and
measurements explanations would unify the mechanics of the shrink fits internal friction
and the internal friction due to the material hysteresis.
83
CHAPTER VI
ROTORDYNAMIC MODELING USING XLTRCTM
XLTRCTM is a Microsoft Excel based rotordynamic software. It presents the
analysis results and inputs in Excel worksheets and is operated from Microsoft Excel.
The code used to perform the rotordynamic analysis is finite element based (FEM). In
this code, a rotor system for which the rotordynamic analysis is required is modeled as
an assemblage of a finite number of elements (which explains the name of the method,
the finite element method), with specified geometry and material elastic properties. The
code assembles the system matrices and performs various rotordynamic analyses, such
as free-free modes, damped eigenvalues, undamped critical speeds, unbalance response
plots against rotating speeds (Bode plots), orbits at any given speed, and transient
analysis. The XLTRCTM software allows users to specify the bearing connection from
the rotor to the ground, or a bearing connection from one shaft to the other shaft with
any stiffness and damping coefficients for the bearing that the user wants to apply for a
particular problem at hand. The XLTRCTM suit also includes various built-in files for
computation of rotordynamic coefficients of hydrodynamic bearings. In these files, the
user specifies the geometrical and tribological properties of the bearings and the
software computes the bearing rotordynamic coefficients. These bearings can then be
“connected” to the rotor, such that the rotor is simulated as if it is mounted on some
particular hydrodynamic bearing. In addition, there are ball bearing codes that compute
the rotordynamic coefficients for any given configuration and then that file could be
connected to the rotor model to model the rotor mounted on given ball bearings. Another
application of user-defined bearing files is to connect different parts of a rotor to model
internal friction. For modeling the internal friction due to shrink fit and interference fits,
the internal friction parameters can be specified in the form of direct moment stiffness
and damping coefficients at an interface of a rotor. The concept is illustrated in more
detail in the following pages when the rotor modeling with shrink fit interface is
described.
84
OVERVIEW OF MODELING USING XLTRCTM
The construction of a rotordynamic model in XLTRC is explained in a step- by -
step manner as follows:
(a) Rotor model
In XLTRCTM, there is a worksheet called the “Model” worksheet. This is the
input worksheet. It is a highly important worksheet, because all the modeling starts from
here. In this worksheet, the user specifies the material and geometry of various elements
that will constitute the rotor. Often times, it happens that the model is a complicated one
such that it requires various rotating parts that are either separate material or such that
some parts that are non-rotating. In instances like these, XLTRCTM provides the users
with the option of specifying multiple shafts within the model.
There are three sections in the “Model” worksheet.
(i) Shafts
In this section, the user specifies the number of shafts that will be used in the
model. The shafts are labeled as 1, 2….etc. In addition, the user specifies the Cartesian
coordinates of the starting points of the various shafts. Also, the user has to specify
whether the shaft is rotating or if it is a stationary shaft. To include material hysteresis,
there is an option for specifying the coefficient of hysteresis. Finally, a column in this
section is “Whirl to Spin Ratio”. This means the user has to specify whether a particular
shaft, for steady state motion, will be executing the synchronous whirl (in which case the
ratio should be equal to 1) or asynchronous whirl (in which case the ratio is different
from 1).
(ii) Material properties
In this section, the user specifies the elastic properties of the materials that will
constitute the rotor model. The user can specify multiple materials. The material
85
properties that need to be inputted are the weight density, Young’s elastic modulus, the
shear modulus and the shear constant. The most common material that is used in
industry to build turbomachinery is steel. There are various grades of steel that are used
in industry, such as AISI 4140 or AISI 4340. However, their material properties are very
similar. Thus, as is the most common case, if the material employed is steel for the rotor,
then the material properties for steel are : weight density = 0.283 lb/in3, elastic modulus
= 30 x 106 psi, and shear modulus = 12 x 106 psi. The shear constant is generally taken
to be zero for Bernoulli-Euler beam model, whereas it is taken to be -1 for Timoshenko
beam model (in which case the effects of rotary inertia are also considered). Apart from
steel, the user can specify any different material properties for any other metal or
material to construct the appropriate models.
(iii) Shaft elements
This section of the worksheet is where the user specifies the geometry of the
elements and associates them with the material properties. This is where building of the
rotor model begins, based on information provided in (i) and (ii) above. In this section,
the user specifies the shaft number, the element number, the sub-element number, the
layer, the length of the element, the outside diameter of the left end of the element, the
inside diameter of the left element, the outside diameter of the right end of the element,
the inside diameter of the right end, the shear interaction factor (0 or 1), and the
associated material with each element.
Associated with each shaft are the elements. Elements are the building blocks of
a rotor shaft. For example, consider the modeling of a cylindrical rotor of uniform
outside diameter, which is 24 inches in length. Finite element method requires that the
domain of interest (in this case, the rotor shaft) be divided into a number of elements so
that corresponding to each element, the calculations could be carried out and then the
system matrices are assembled together to perform various rotordynamic calculations of
interest. So suppose that the 24 inches shaft is divided into 12 equal length pieces. These
twelve pieces are the elements. However, the elements need not be of the same length,
86
neither they need to be made of the same material. Also, the total number of elements
used to model the rotor can be different from twelve. As a rule, the larger the number of
elements, the more accurate the analysis is, but the penalty associated with higher
number of elements is the larger calculation time. For simple enough rotor systems such
as the cylindrical shaft of uniform outside diameter, it is unnecessary to divide it into a
large number of elements, when a reasonably well analysis could be carried out with
smaller number of elements.
Horizontal (axial) sub-division of an element is called as Sub-Element. Similarly,
radial sub-division of an element is called a Layer. Accordingly, when sub-elements for
an element are specified in “Model” worksheet, then the element number corresponding
to each of them is the same, but in the sub-element column, they are specified as
1,2…etc. Similarly, when layers are specified for an element, then the element number
corresponding to the two layers will be the same, but in the layer column, they will be
labeled as 1,2…etc. For layers, the length should be the same, because they belong to the
same element. But the outside and inside diameters for layers need not be same.
However, for sub-elements, the lengths need not be same. Their individual lengths sum
will determine the total length of that particular element.
In this way, the rotor geometry is constructed. This is the most basic and
important step towards the rotordynamic analysis. Once the geometry is developed, it
can be viewed as a front view (which means in a plane) using the “Geo Plot” tool in
XLTRCTM. This tool can be accessed through the XLTRCTM toolbar or through the
XLTRC drop-down menu. Viewing the “Geo Plot” worksheet shows user how the rotor
model looks like. It may give a clue to the user to detect any discrepancy to note in the
geometry that may not agree with how the model is supposed to look like. In other
words, viewing the geometric plot is a good practice to check for possible geometric
errors in the “Model” worksheet, besides checking the model geometry data entries.
87
CONSTRUCTION OF A SINGLE-DISK ROTOR MODEL
Construction of a single-disk rotor model will be explained in a detailed and step-
by-step manner as follows:
(a) Rotor model
Fig. 47 Finite element model of a single-disk rotor using XLTRCTM
Fig.47 shows the view of the single-disk model as viewed through Geo Plot tool
using XLTRCTM. It can be seen from Fig.47 that the rotor is modeled as a system of
finite elements of the main rotor shaft (Shaft 1) and the disk (Shaft 2). The station
numbers are numbered consecutively from 1 (at the far left position, for shaft 1) to 13
(far right position of Shaft 1) and 14 to 16 (the elements constituting the disk, which is
considered as Shaft 2). In figure 1, each horizontal division of a shaft constitutes one
element. Therefore, there are 12 elements of Shaft 1, while there are two elements of
Shaft 2. Each element has a left-end station number and a right-end station number. In
figure 1, for example, associated with Shaft 1, element 1 has left end station number as 1
Shaft 1
1 2 3 4 5 6 7 8 9 10 11 12
14 15 16
13
Shaft 2
88
and right-end station number as 2. Similarly, for Shaft 1, element 2 has left-end station
number 2 and right-end station number 3, and so on. It should be emphasized here that
since the interest is to model internal friction that occurs between the main rotor shaft
and the disk at the interface, the main shaft and the disk are modeled as separate shafts in
XLTRCTM. If these were modeled as one shaft, then it would be necessary to specify the
“Layer” option and the appropriate nomenclature for the layers in the “Model”
worksheet. However, it would then not be possible to model the internal friction,
because the latter model would mean physically a single shaft and will assume no
interaction between the main rotor shaft and the disk. Therefore, to model internal
friction, it is necessary to model this system as a system containing two shafts (Shafts 1
and 2) and then specifying some internal friction parameters at the interface of Shafts 1
and 2.
(b) Bearing support connection
Once the rotor model is constructed, it is then necessary to specify the bearings
that support the rotor to initiate some useful rotordynamic analysis on it. Most often, it is
necessary to begin the analysis by first finding out the free-free mode shapes and natural
frequencies of the rotor. To accomplish this, the user can either: (1) specify the zero
mass, stiffness and damping bearings that connect the two rotor ends to the ground and
then running the damped eigenvalue worksheet, or (2) without having specified the zero
bearings attached to the rotor, the user can directly use the “Free-Free Modes” command
from XLTRCTM pull-down menu and it will compute the free-free modes values for the
user.
Once the free-free analysis is accomplished, however, it is necessary for further
analysis that the user specifies a non-zero bearing connection from the rotor to the
ground. For analysis on the single-disk rotor, the bearings that were used were the ball
bearings. The rotordynamic coefficients files for the ball bearings of various
configurations can be obtained through XLTRCTM suit. Once the bearing specifications
are described and the bearing coefficients are obtained, the title of that file is copied.
89
Then the rotor model file is opened and then that title head is “Paste Special” to the right
most column, with heading “Connection” in “Brg” worksheet in the rotor model file.
The result for this operation will look as follows:
Fig. 48 Rotor model with bearing connections, connecting the rotor with the ground
In Fig.48, the bearing connections are from stations 2 and 11 to the ground. In
XLTRCTM, this will be specified in the “Brg” worksheet file as connecting station 2 to
station 0 and station 11 to station 0. In XLTRCTM, station “0” is always considered as
ground with no exception. Now the model shown in figure 2 can be used to evaluate
several rotordynamic results of interest such as response plots against speed due to
14 15 16
Shaft 2
Shaft 1
1 2 3 4 5 6 7 8 9 10 11 12
90
imbalance, the damped eigenvalues, the undamped critical speeds and the transient
analysis. The model shown in Fig.41 will be suitable to model a rotor that has the
bearings mounted on very stiff (or almost rigid) foundations, such that the foundation
flexibility effect can be ignored. However, for foundations showing flexibility and
motion in either or both horizontal and vertical direction need to be modeled as a
separate shaft in the model and connected to the rotor model through the bearings
coefficients, while to the grounds, they should be connected through their modal
parameters.
(c) Foundation effect
As stated, the foundation effect can be incorporated by connecting the foundation
mass to the bearings and the ground through the foundation modal parameters, such as
stiffness, mass and damping through user-defined file in XLTRCTM, which is called
XLUseKCM. In the simulation of single-disk rotor, the foundation modal parameters
were determined through experimental modal testing on the foundation using a
calibrated modal hammer and an accelerometer. When the foundation modal parameters
are established, they can then be inputted into a user-defined stiffness, mass and
damping file. This file is then connected to the rotor through bearing connection, and it
is also connected to the ground. When such operations are performed, then the Geo Plot
tool in XLTRCTM shows the modified model as shown in Fig. 49:
91
Fig. 49 Rotor model with bearings and foundation included
Fig. 49 shows the foundation included with the rotor model. The foundation is
modeled as Shaft 3, which is non-rotating. It is seen from Fig.49 that the foundation is
connected to the bearings at the station numbers 18 and 27. Thus the rotor is connected
to the bearings through stations 2 and 11 and the bearings are connected to the
Shaft 1
1 2 3 4 5 6 7 8 9 10 11 12
14 15 16
Shaft 2
Shaft 3
17 18 19 20 21 22 23 24 25 26 27 28 29
92
foundation through 2 to 18 and 11 to 27. Then the foundation modal parameters are
connected to the ground through 18, 23 and 27. Thus, in XLTRCTM, the connections will
be as: stations 2 to 18, ball bearing connection; stations 11 to 27, ball bearing
connection; stations 18 to 0, user-defined (and experimentally obtained) foundation
modal parameters and similarly, for stations 23 to 0 and 27 to 0, foundation modal
parameters will be the connecting file.
(d) Modeling of internal friction
In order to model the internal friction that acts at the interface of Shafts 1 and 2
(the main rotor shaft and the disk, respectively), use is made of a user-defined moment
coefficients file, which is called as XLUseMoM in XLTRCTM. In this file, the internal
friction forces and moments (in the plane of the disk) can be modeled by inputting
various stiffness and damping coefficients. For forces, there are direct and cross-coupled
stiffness coefficients as well as the damping coefficients. For moments, there are
similarly direct and cross-coupled moment coefficients. Once the inputs are established,
then the title of the file is copied and pasted to the “Brg” worksheets, under the column,
“Connection”. In establishing the internal friction moments, it should be noted from Fig.
3 that the connection will occur between stations 6 to 14 and 8 to 16. That is, the three
points at the interface of Shaft 1 and Shaft 2. This is illustrated in Fig.50 below:
93
Fig. 50 Interface points of Shafts 1 and 2, where internal friction parameters are
specified. The points are connected through a user-defined moment coefficients file.
After completing the steps to model the rotor, the model can be analyzed using
various options to simulate rotordynamics of the rotor. To analyze the stability, the
‘EIG’ worksheet is run to evaluate the damped eigenvalues of the rotor-bearing system.
A negative damped eigenvalue (damping ratio) indicates an unstable mode. The
eigenvalues are generated in XLTRCTM as pair. The real part of an eigenvalue is the
damping ratio, whereas imaginary part is the damped natural frequency. Therefore, both
the stability and frequency of mode are calculated using the “EIG” tool in XLTRCTM.
6 7 8
Shaft 1
Shaft 2
14 15 16
Interface points
94
Fig. 51 Internal moments acting on the rotor disks (in its plane of deflection) in X and Y directions and their resultant moment vector, MR. The resultant moment tends
to bend the shaft in the direction of forward whirl.
The shrink fit between the shaft and the disk is modeled in XLTRCTM using
XLUseMoM file. Physically, the shrink fit is modeled as if there are linear translational
and torsional springs and dampers between the sleeve and the disk interface.
In XLTRCTM, there is no force that can be specified internal or external to the
model in axial (Z) direction. Physically, the internal friction moment acts so as to
develop a forward whirl of the rotor (Fig.51). With the help of XLUseMoM file, the
internal friction moments are modeled as internal moments to the system i.e., no reaction
acting on the ground (which is an incorrect approach).
XLTRCTM models the internal forces and moments based on the following
equation:
MR
MX
MY
+Ф
X
-MX
-MY
Y
95
xx xy xax xay xx xy xax xayX
yy yx yax yay yx yy yax yayY
X xax xay axax axay xax xay axax axay
Y yax yay ayax ayay yax yay ayax ayay
XK K K K C C C CF XK K K K C C C CF Y Y
M K K K K C C C CM K K K K C C C C
αα
β
•
•⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥= − −⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎣ ⎦⎣ ⎦ ⎣ ⎦ ⎣ ⎦
β
•
•
⎡ ⎤⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥⎢ ⎥⎣ ⎦
(60)
From equation (60) above, it can be seen that the forces and the moments acting
as internal loadings in the system can be modeled with the specified stiffness and
damping coefficients. In particular, the moments MX and MY can be modeled. As shown
in Fig.51 and Fig.52, the coefficients can be so selected such that these moments act to
bend the shaft in the direction of forward whirl.
96
Fig. 52 XLUseMoM Worksheet for entering of internal friction parameters
97
CHAPTER VII
ROTORDYNAMIC SIMULATIONS OF EXPERIMENTS USING THE
INTERNAL MOMENTS MODEL
As discussed in Chapter IV, the cross-coupled moment expressions for various
interface friction models contain de-stabilizing terms. The effect of these cross-coupled
terms is to add energy to the rotor-bearing system above the first critical speed and to
dissipate energy from the system below the first critical speed. As described in Chapter
III, the internal moment model is the physically correct model for interface friction in
shrink fit joints. This chapter details the application of the internal moment model using
XLTRCTM rotordynamics software to the rotordynamics and stability study of the
experimental rotor-bearing systems with shrink fit and interference fit interfaces, as
described in Chapter II. The results of the simulations show that the internal moment
model can be used to predict the stability of the experimental setups without having the
need to use physically incorrect follower force model. The simulation results from this
chapter are the first step towards the application of a physically correct model of the
shrink fit interface friction. The simulation results from this chapter also provide a
motivation for an advanced research on the determination of the unknown interface
friction parameters such as direct moment stiffness and damping coefficients that can be
used to determine the threshold speeds of instability of a rotor-bearing system with
shrink fit interfaces, rather than using the physically incorrect follower force model
(Gunter’s model) to determine the stability of a rotor-bearing system.
The simulations of various experimental rotor setups with shrink fit interfaces are
described in the following pages.
98
SINGLE-DISK ROTOR SIMULATIONS
The single-disk rotor has an interference fit interface through a tapered sleeve
between the disk and the shaft, as described in Chapter II. In modeling the single-disk
rotor using XLTRCTM, the internal moments are specified as the moment coefficients
between the disk and the shaft, thereby modeling the interface friction. The interface is
shown in Fig.53:
Fig. 53 Close-up view of the single-disk rotor showing the shaft and disk interface
through tapered sleeve.
The single-disk rotor model with the foundation is shown in Fig.54:
Sliding interface of disk and shaft
Shaft
Disk
Tapered sleeve
99
Fig. 54 Single-disk rotor simulation using XLTRCTM
Fig.54 shows the model of the single-disk rotor which is analyzed using
XLTRCTM software. In modeling the single-disk rotor-bearing system, the system is
modeled as three shafts. ‘Shaft 1’ is the steel shaft with a uniform diameter of 2.5 inches,
and 52.50 inches long. ‘Shaft 2’ is the steel disk with an outside diameter of 10 inches,
Shaft 1
1 2 3 4 5 6 7 8 9 10 11 12
Shaft 2
Shaft 3
14 15 16
100
and an axial length of 5 inches, mounted at the mid-span of ‘Shaft 1’, which is the steel
shaft. ‘Shaft 3’ is the foundation, which does not participate in rotation and whirling, but
vibrates in horizontal and vertical directions due to its stiffness and inertia properties.
To simulate the interface friction between the disk (Shaft 2) and the shaft (Shaft
1), the internal moment bearing file is used to make a connection between the two shafts.
In the experimental setup for the single-disk rotor as described in Chapter II, the tapered
sleeve that creates the interference fit between the uniform shaft and the disk acts such
that the relative sliding between the shaft and the disk during whirling motion of the
rotor takes place at the end which is inside the disk. The other end of the tapered sleeve
which is used to fasten the sleeve to the disk through the push bolts is less likely to slide
relative to the shaft due to higher interface pressures at that location.
With this reasoning, the internal moment bearing connection is applied between
the points 6 and 14, whereas points 7-15 and 8-16 are constrained through kinematical
constraint of executing the same translational and rotational motion. In other words, the
sliding is specified at the interface 6-14, but no sliding is specified between other points
on the interface.
The foundation which is modeled as ‘Shaft 3’ is connected to the rotor through
ball bearing files and to the ground through user-defined stiffness and damping
coefficients. The user-defined stiffness and damping coefficients are estimated using
modal tests on the foundation in the horizontal and vertical directions.
1. Loose fit simulation
In simulating the rotordynamic instability of the single-disk rotor with looser fit
due to which the rotor-bearing system became unstable at 5800 rpm, the interface
coefficients were determined by trial and error until the simulation results matched with
the experimental results. In the experiments, the single-disk rotor became unstable at a
threshold speed of 5800 rpm, with the unstable sub-synchronous vibrations occurring at
the first eigenvalue of the rotor-bearing system, which is around 3000 cpm.
101
Table 1. Coefficients of the internal moment applied at the interface of the shaft and the disk for loose fit.
Speed KXX KYY CXX CYY KaXaX KaXaY KaYaX KaYaY CaXaX CaYaY
RPM Lb/in Lb/in Lb-s/in Lb-s/in Lb-in Lb-in Lb-in Lb-in Lb-in-s Lb-in-s
0 8.5e+8 8.5e+8 1000 1000 1e+4 0 0 1e+4 6923.2 6923.2
1000 8.5e+8 8.5e+8 1000 1000 1e+4 7.25e+5 -7.25e+5 1e+4 6923.2 6923.2
2000 8.5e+8 8.5e+8 1000 1000 1e+4 1.45e+6 -1.45e+6 1e+4 6923.2 6923.2
3000 8.5e+8 8.5e+8 1000 1000 1e+4 2.18e+6 -2.18e+6 1e+4 6923.2 6923.2
4000 8.5e+8 8.5e+8 1000 1000 1e+4 2.90e+6 -2.90e+6 1e+4 6923.2 6923.2
5000 8.5e+8 8.5e+8 1000 1000 1e+4 3.63e+6 -3.63e+6 1e+4 6923.2 6923.2
6000 8.5e+8 8.5e+8 1000 1000 1e+4 4.35e+6 -4.35e+6 1e+4 6923.2 6923.2
7000 8.5e+8 8.5e+8 1000 1000 1e+4 5.08e+6 -5.08e+6 1e+4 6923.2 6923.2
8000 8.5e+8 8.5e+8 1000 1000 1e+4 5.80e+6 -5.80e+6 1e+4 6923.2 6923.2
9000 8.5e+8 8.5e+8 1000 1000 1e+4 6.53e+6 -6.53e+6 1e+4 6923.2 6923.2
10000 8.5e+8 8.5e+8 1000 1000 1e+4 7.25e+6 -7.25e+6 1e+4 6923.2 6923.2
11000 8.5e+8 8.5e+8 1000 1000 1e+4 7.98e+6 -7.98e+6 1e+4 6923.2 6923.2
12000 8.5e+8 8.5e+8 1000 1000 1e+4 8.70e+6 -8.70e+6 1e+4 6923.2 6923.2
102
Table 1 shows the user-defined interface friction parameters that are applied
between points 6 and 14 in the single-disk rotor to model to simulate the rotordynamic
instability due to slip at the shaft-disk interface. The cross-coupled stiffness coefficients
(KaXaY, KaYaX) are functions of the rotational speed, as discussed in Chapter IV. As
shown in Chapter IV, the cross-coupled moment coefficients are related to the direct
moment damping coefficients (CaXaX, CaYaY) with the following equations:
ωaXaYaXaY CK = (61)
ωaYaXaYaX CK −= (62)
In equations (61) and (62), the rotational speed is expressed in the units of
radians per second (rad/s) rather than rotations per minute (rpm).
The remaining stiffness and damping coefficients which are not shown in Table 1
are assumed to be zero. From Table 1, the value of the interface moment damping value
is 4060 lb-in-s/rad, and the direct moment stiffness coefficient is 1e+04 lb-in/rad. The
high values of the direct translational stiffness and damping coefficients (KXX, KYY, CXX,
CYY) are indicative of relatively tight connection between the points 6 and 14, thereby
indicating that the relative translational motion in X and Y directions is negligible at the
interface. However, the values of the direct moment stiffness and damping coefficients
(KaXaX, KaYaY, CaXaX, CaYaY) are indicative of how easily the shaft can slide independent
of the disk. The simulated eigenvalues using the coefficients indicated in Table 1 show
that the threshold speed of instability of the rotor-bearing system is 5800 rpm, with the
unstable mode frequency as the first eigenvalue of the rotor-bearing system, which is
3100 cpm.
An output from XLTRCTM in Fig. 52 shows the unstable mode shape of the
rotor-bearing system above the threshold speed of instability:
103
Damped Eigenvalue Mode Shape PlotSingle-disk rotor
Modeling internal friction using internal moments
f=3168.2 cpmd=-.0008 zetaN=6400 rpm
forwardbackward
Fig. 55 Unstable mode shape of the single-disk rotor above the threshold speed
Fig.55 shows the unstable mode shape of the single-disk rotor evaluated using
XLTRCTM. The speed of rotation is 6400 rpm which is above the threshold speed of
instability (5800 rpm). The frequency of the unstable mode is 3168 cpm, which is equal
to the first eigenvalue of the rotor-bearing system. The value of the damping ratio is
-0.0008 which shows that the mode is unstable. Fig.55 shows that the unstable mode
shape involves bending of the shaft and no motion from the foundation. This is the most
favorable condition for instability due to shrink fit to occur. In this mode of shaft
bending only, the relative sliding between the disk and the shaft will create friction
forces (and the equivalent friction moments) at the interface and initiate instability. In
addition, the mode shape shows that although the shaft whirls with elliptical orbits, the
orbits are not totally planar, and therefore energy can be added to the orbits to initiate
instability. The mode shape is a function of several factors, such as foundation stiffness
Foundation (‘Shaft 3’)
104
asymmetry, amount of internal friction (as specified in Table 1) and rotor-bearing
physical parameters.
2. Tight fit simulation
As shown in Chapter II, when the interference fit between the disk and the shaft
is increased (radially around 2.2 mils), the threshold speed of instability of the single-
disk rotor is increased to around 11000 rpm, as compared to 5800 rpm when the fit is
loose (1.16 mils radially). This section discusses the simulation of the single-disk rotor
with the tight fit. The simulations show that the cross-coupled moment stiffness
coefficients required to predict the threshold speed of instability of about 11000 rpm are
smaller in magnitude as compared to the coefficients required to predict the threshold
speed of instability of 5800 rpm. The simulations also show that the values of the direct
moment stiffness coefficients are larger as compared to the case of loose fit. The direct
moment stiffness coefficients indicate how easily the shaft can slide relative to the sleeve
and disk. Larger value of the direct stiffness coefficients will mean that the shaft will not
slide relative to the sleeve with the same ease in case of tight fit (due to high interface
pressures) as it slid relative to the sleeve when the interference fit was relatively small
(resulting in lower interface pressures). This physically means that increasing the
tightness of the fit corresponds to lesser slip between the sleeve and the shaft (due to
high interface pressures) and therefore decreased amount of internal moments at the
interface. As shown in Chapter IV, the magnitude of friction moments for viscous
friction model is proportional to the amplitude of micro-slip at the interface. The
amplitude of micro-slip decreases with the increase of interference fit, which
corresponds with the higher values of direct moment stiffness coefficients and lower
values of the micro-slip amplitudes. The internal moment coefficients required to
simulate the experimental setup of the single-disk rotor with the tight fit is shown in
Table 2 on the next page:
105
Table 2. Coefficients of the internal moment applied at the interface of the shaft and the disk for tight fit.
Speed KXX KYY CXX CYY KaXaX KaXaY KaYaX KaYaY CaXaX CaYaY
RPM Lb/in Lb/in Lb-s/in Lb-s/in Lb-in Lb-in Lb-in Lb-in Lb-in-s Lb-in-s
0 8.5e+8 8.5e+8 1000 1000 1e+5 0 0 1e+5 2148.6 2148.6
1000 8.5e+8 8.5e+8 1000 1000 1e+5 2.25e+5 -2.25e+5 1e+5 2148.6 2148.6
2000 8.5e+8 8.5e+8 1000 1000 1e+5 4.50e+5 -4.50e+5 1e+5 2148.6 2148.6
3000 8.5e+8 8.5e+8 1000 1000 1e+5 6.75e+5 -6.75e+5 1e+5 2148.6 2148.6
4000 8.5e+8 8.5e+8 1000 1000 1e+5 9.0e+5 -9.0e+5 1e+5 2148.6 2148.6
5000 8.5e+8 8.5e+8 1000 1000 1e+5 1.125e+6 -1.125e+6 1e+5 2148.6 2148.6
6000 8.5e+8 8.5e+8 1000 1000 1e+5 1.35e+6 -1.35e+6 1e+5 2148.6 2148.6
7000 8.5e+8 8.5e+8 1000 1000 1e+5 1.575e+6 -1.575e+6 1e+5 2148.6 2148.6
8000 8.5e+8 8.5e+8 1000 1000 1e+5 1.80e+6 -1.80e+6 1e+5 2148.6 2148.6
9000 8.5e+8 8.5e+8 1000 1000 1e+5 2.025e+6 -2.025e+6 1e+5 2148.6 2148.6
10000 8.5e+8 8.5e+8 1000 1000 1e+5 2.25e+6 -2.25e+6 1e+5 2148.6 2148.6
11000 8.5e+8 8.5e+8 1000 1000 1e+5 2.475e+6 -2.475e+6 1e+5 2148.6 2148.6
12000 8.5e+8 8.5e+8 1000 1000 1e+5 2.70e+6 -2.70e+6 1e+5 2148.6 2148.6
106
Damped Eigenvalue Mode Shape PlotSingle-disk rotor
Modeling internal friction using internal moments
f=2894.5 cpmd=-.0006 zetaN=11400 rpm
forwardbackward
Fig. 56 Unstable mode shape of the single-disk rotor above the threshold speed (for the case of tight fit)
Fig.56 shows the unstable mode plot using XLTRCTM for the single-disk rotor
model with the tight fit case. The threshold speed of instability for the tight fit is 11000
rpm and the mode shape in Fig.56 is plotted at the speed of 11400 rpm. As in the case of
loose fit (Fig.55), the mode shape shows that the rotor shaft bends but the foundation
does not vibrate in this mode. The mode shapes are not totally planar, which shows that
the energy can be added to the system through rotor motion and the system has a
tendency to grow in unstable motion. The unstable mode frequency is 2900 cpm, which
is the same as the experimentally determined frequency of the unstable sub-synchronous
whirl motion. The damping ratio is -0.0006, which shows that the mode is unstable.
107
TWO-DISK ROTOR SIMULATIONS
The internal friction moments model is applied to the rotordynamic simulations
of the two-disk rotor setup as discussed in Chapter II. As noted in Chapter II, several
configurations of the two-disk rotor with different sleeve geometries and shrink fits were
experimentally tested, of which only one configuration is unstable. This section will
demonstrate the application of the internal moments model to predict the rotordynamic
stability of experimentally tested two-disk rotor configurations. An isometric view of the
two-disk rotor with various internal features visible is shown in Fig.57:
Fig. 57 Isometric view of the two-disk rotor showing internal features
Stepped-section steel shaft
Steel disks having shrink fit interfaces with aluminum sleeve
Hollow aluminum sleeve
Thin (2 in. dia) section of steel shaft
108
1. Tight fit at one end, loose fit at the other end
In this configuration, the aluminum sleeve has an outside diameter of 10 inches.
The sleeve has an undercut at one of the ends (see Chapter II, put picture in Chapter II of
undercut sleeve). The undercut end has 5 mils diametral shrink fit with the steel disk,
whereas the other interface of the sleeve and disk have 12 mils diametral shrink fit. As
shown in the experimental results in Chapter II, this rotor configuration became unstable
at a threshold speed of instability of 9600 rpm, and eventually wrecked during the coast-
down. The internal moments model is applied to the two-disk rotor model to
demonstrate the instability of this rotor configuration. In modeling the shrink fits at the
two ends, the tight fit end is made to have high values of direct moment stiffness
coefficients with zero cross-coupling stiffness coefficients to simulate the condition of
very tight fit (with zero slip and friction). The other end with the undercut is connected
to the steel disks in the rotor model through internal moment coefficients that have
relatively lower values of direct stiffness coefficients and high values of cross-coupled
moment stiffness coefficients as compared to the tight fit end to simulate the relative slip
between the disk and the sleeve at that end. In addition, the modal parameters of the
foundation connecting to the ground (stiffness and damping) for the two-disk rotor setup
are input in XLTRCTM. The stiffness and damping coefficients of the foundation for the
two-disk rotor setup are different from the ones for the single-disk rotor setup, because
an additional stiffener was installed on the foundation in an attempt to reduce stiffness
asymmetry for the experiments involving the two-disk rotor. These coefficients are
described in Appendix C. The coefficients of the internal moment bearing file for each
of the two ends are shown on next pages in Tables 3 and 4:
109
Table 3. Coefficients of internal moment applied at the undercut end for two-disk rotor simulation.
Speed KXX KYY CXX CYY KaXaX KaXaY KaYaX KaYaY CaXaX CaYaY
RPM Lb/in Lb/in Lb-s/in Lb-s/in Lb-in Lb-in Lb-in Lb-in Lb-in-s Lb-in-s
0 8.5e+8 8.5e+8 1000 1000 1e+4 0 0 1e+4 2150 2150
1000 8.5e+8 8.5e+8 1000 1000 1e+4 2.25e+5 -2.25e+5 1e+4 2150 2150
2000 8.5e+8 8.5e+8 1000 1000 1e+4 4.50e+5 -4.50e+5 1e+4 2150 2150
3000 8.5e+8 8.5e+8 1000 1000 1e+4 6.75e+5 -6.75e+5 1e+4 2150 2150
4000 8.5e+8 8.5e+8 1000 1000 1e+4 9.00e+5 -9.00e+5 1e+4 2150 2150
5000 8.5e+8 8.5e+8 1000 1000 1e+4 1.13e+6 -1.13e+6 1e+4 2150 2150
6000 8.5e+8 8.5e+8 1000 1000 1e+4 1.35e+6 -1.35e+6 1e+4 2150 2150
7000 8.5e+8 8.5e+8 1000 1000 1e+4 1.58e+6 -1.58e+6 1e+4 2150 2150
8000 8.5e+8 8.5e+8 1000 1000 1e+4 1.80e+6 -1.80e+6 1e+4 2150 2150
9000 8.5e+8 8.5e+8 1000 1000 1e+4 2.03e+6 -2.03e+6 1e+4 2150 2150
10000 8.5e+8 8.5e+8 1000 1000 1e+4 2.25e+6 -2.25e+6 1e+4 2150 2150
11000 8.5e+8 8.5e+8 1000 1000 1e+4 2.48e+6 -2.48e+6 1e+4 2150 2150
12000 8.5e+8 8.5e+8 1000 1000 1e+4 2.70e+6 -2.70e+6 1e+4 2150 2150
110
Table 4. Coefficients of internal moment applied at the tight fit end for two-disk rotor simulation.
Speed KXX KYY CXX CYY KaXaX KaXaY KaYaX KaYaY CaXaX CaYaY
RPM Lb/in Lb/in Lb-s/in Lb-s/in Lb-in Lb-in Lb-in Lb-in Lb-in-s Lb-in-s
0 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
1000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
2000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
3000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
4000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
5000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
6000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
7000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
8000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
9000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
10000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
11000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
12000 8.5e+8 8.5e+8 1000 1000 5e+8 0 0 5e+8 0 0
111
Damped Eigenvalue Mode Shape PlotShell Rotor test rig
Modeling the Interference f it
f=6570.2 cpmd=-.0014 zetaN=11000 rpm
forwardbackward
Fig. 58 Unstable mode shape of two-disk rotor model with different fits at two interfaces
Table 3 shows the internal moment coefficients at the loose fit end, whereas
Table 4 shows the internal moment coefficients at the tight fit with no cross-coupling.
Fig.58 shows the mode shape plot for the two-disk rotor model at a speed of
11000 rpm. The plot is generated using the XLTRCTM software. The mode shape shows
that the unstable whirling frequency is 6570 cpm (109.5 Hz) which was one of the
unstable modes excited and measured in the experimental results for the two-disk rotor
with different fits at two interfaces. The damping ratio is -0.0014, which shows that the
mode is unstable.
112
2. Same fit at both ends
To simulate same tightness of fit at both ends, the internal moment coefficients at
both ends are provided with the same direct moment stiffness and damping coefficients.
The experimental setup with the same tightness of fits is discussed in Chapter II. In this
experimental setup for the two-disk rotor, the outside diameter of the aluminum sleeve
was 9.5 inches in one setup, and 9.25 inches in another setup. In both cases, the shrink fit
was the same at both the disks (10 mils diametrally). The experimental results showed
that the rotor-bearing system did not become unstable in either of those configurations,
when the shrink fit was the same at both the ends. Even though in one of the
configurations, the sleeve was heated at around 11000 rpm until the shrink fit contact
was lost, still no instability was observed in the rotor-bearing system. The possible
reason that the rotor-bearing system did not become unstable in the symmetric fit
configurations may be because of sleeve’s bending together with the steel shaft (due to
smaller outside diameters, that is, 9.5 inches or 9.25 inches as compared to 10 inches
outside diameter for unstable configurations), due to which no appreciable relative
sliding took place and no friction is generated. The other reason may be an inherently
small amount of direct damping in the joint for these particular configurations, which
maybe a function of sleeve’s geometry and shrink fit values.
The application of internal moments model to the case of same fit at both the
disks shows that for some choice of direct moment stiffness and damping coefficients,
the two-disk rotor bearing system does not show instability. This indicates that the
internal moment coefficients are functions of sleeve geometry and shrink fit values. A
high value of the shrink fit corresponds to corresponding high values of the direct
moment stiffness coefficients (KaXaX, KaYaY). In addition to this, the direct moment
damping coefficients (CaXaX, CaYaY) are also possibly the functions of sleeve geometry
are the shrink fit values.
The coefficients of the internal moment are presented in Table 5 on the next
page:
113
Table 5. Coefficients of the internal moment applied at the two interfaces for the same fit case of the two-disk model.
Speed KXX KYY CXX CYY KaXaX KaXaY KaYaX KaYaY CaXaX CaYaY
RPM Lb/in Lb/in Lb-s/in Lb-s/in Lb-in Lb-in Lb-in Lb-in Lb-in-s Lb-in-s
0 8.5e+8 8.5e+8 1000 1000 1e+4 0 0 1e+4 95.49 95.49
1000 8.5e+8 8.5e+8 1000 1000 1e+4 1e+4 -1e+4 1e+4 95.49 95.49
2000 8.5e+8 8.5e+8 1000 1000 1e+4 2e+4 -2e+4 1e+4 95.49 95.49
3000 8.5e+8 8.5e+8 1000 1000 1e+4 3e+4 -3e+4 1e+4 95.49 95.49
4000 8.5e+8 8.5e+8 1000 1000 1e+4 4e+4 -4e+4 1e+4 95.49 95.49
5000 8.5e+8 8.5e+8 1000 1000 1e+4 5e+4 -5e+4 1e+4 95.49 95.49
6000 8.5e+8 8.5e+8 1000 1000 1e+4 6e+4 -6e+4 1e+4 95.49 95.49
7000 8.5e+8 8.5e+8 1000 1000 1e+4 7e+4 -7e+4 1e+4 95.49 95.49
8000 8.5e+8 8.5e+8 1000 1000 1e+4 8e+4 -8e+4 1e+4 95.49 95.49
9000 8.5e+8 8.5e+8 1000 1000 1e+4 9e+4 -9e+4 1e+4 95.49 95.49
10000 8.5e+8 8.5e+8 1000 1000 1e+4 1e+5 -1e+5 1e+4 95.49 95.49
11000 8.5e+8 8.5e+8 1000 1000 1e+4 1.1e+5 -1.1e+5 1e+4 95.49 95.49
12000 8.5e+8 8.5e+8 1000 1000 1e+4 1.2e+5 -1.2e+5 1e+4 95.49 95.49
114
The simulations of the experimental results on configurations of the single-disk
and the two-disk rotors provide some values of the internal moment coefficients for
different configurations. The simulations show that the internal moment coefficients
vary from configuration to configuration, and therefore it is challenging to use the
internal moments model to analyze the internal friction and stability of a rotor-bearing
system. In Appendix C, an attempt has been made to analyze the stability of various
imagined rotor configurations (both single-disk and the two-disk) that are different from
the experimental rotor configurations. The imagined rotor models are developed and
analyzed using XLTRCTM software. The internal moments model is used in each of the
imagined configurations to model shrink fit or interference fit interface internal friction.
Although the simulation results are not supported by experimental data, the simulations
nevertheless provide a first step towards modeling, designing and analyzing the
imagined rotor configurations with shrink fit and interference fit interfaces. Another
objective of including the imagined configurations is to assess some stability trend that
can be achieved by varying several rotor elastic, material and geometric properties. See
Appendix C for imagined rotor configurations stability simulations and their discussion.
115
CHAPTER VIII
CONCLUSIONS
The research presented in this dissertation provides an experimental and
theoretical study of the effect of slippage at shrink fit and interference fit interfaces on
the rotordynamic stability of rotor-bearing systems. The experimental study confirms
that the instability caused by the slippage in shrink fit and interference fit joints in a rotor
is potentially catastrophic for rotor-bearing systems. The experimental study also
confirms that the instability caused by internal friction in shrink fit and interference fit
joints is sub-synchronous and usually occurs at the first eigenvalue of the rotor-bearing
system at an operating speed above the first critical speed of the rotor-bearing system.
From the experimental study of the single-disk rotor for steel-steel interface on this
particular rotor setup, it was found that there is a critical value of the interference fit (1
mils radial at speed) that causes the rotor to become unstable above the first critical
speed. It is therefore necessary to have an analysis available that predicts the interference
as a function of speed.
For the two-disk rotor, there is only one configuration out of several tested that
became unstable. In the configuration that became unstable, the shrink fit was very tight
at one end and fairly loose at the other end. Therefore, it seems likely that the loose and
tight distribution of shrink fits along the two-disk rotor interfaces is a significant factor
affecting the stability of the rotor-bearing system.
The theoretical study shows that the internal friction can be modeled more
realistically using the internal friction moments model. However, there are several
variables and unknowns in the internal moments model, such as the direct moment
damping in the joint and the direct moment stiffness coefficients, which make the use of
this model dependent on measurements, such as the experimental value of the threshold
speed of rotordynamic instability. Using the experimental results, the direct moment
damping and stiffness coefficients can be found using XLTRCTM simulations by trial
and error until the simulated results match the experimental values of the first critical
116
speed, the frequency of the unstable mode and the threshold speed of rotordynamic
instability of a rotor-bearing system. This is the procedure that was adopted to compute
the internal friction moment coefficients of the experimental rotors which showed
instability. The same procedure was also used to compute the values of the internal
moment coefficients for the two-disk rotor configurations that did not become unstable.
The experimental results show that the internal moment coefficients vary with different
configurations of the rotor-bearing systems. In the configurations for the two-disk rotor
that did not become unstable during the operating speed range of 0-12000 rpm, the
sleeve outside diameter were 9.5 inches and 9.25 inches, with the same tightness of fits
(about 10 mils diametral on both interfaces, with complete 2 inches axial contact) at the
two interfaces. The simulations of these configurations using the XLTRCTM software
show that the rotor will be stable if the values of the direct moment damping coefficients
are smaller in magnitude as compared to the unstable rotor configuration (for which the
sleeve outside diameter is 10 inches, with a 2 inches axial contact, 12 mils radial fit at
one end and an under-cut 1 inch axial contact, 2.5 mils radial loose fit at the other
interface).
The simulations of the imagined single-disk and two-disk rotor configurations
with shrink fit or interference fit interfaces, though not supported by experimental data,
provide the first step towards assessing stability trends caused by a rotor’s elastic and
geometric properties. The simulations also extend the application of the internal friction
moments model from the experimental rotors to the imagined rotor configurations.
RECOMMENDATIONS FOR FUTURE RESEARCH
For further research on the subject of internal friction caused by slippage in
shrink fit and interference fit joints and their effects on rotordynamic stability, the
parameters involved in the internal moments model such as the direct moment stiffness
and damping coefficients for the viscous friction model, need to be studied in more
detail for predictions of threshold speeds of rotordynamic instability in a rotor-bearing
system. As concluded from the experimental and theoretical study presented in this
117
dissertation, the internal moment coefficients vary from one rotor configuration to
another rotor configuration. In addition, the internal moment coefficients can be
functions of several other system parameters such as rotor mode shape on the bearings,
slope of the mode shape, tightness of fits and geometry of the interface, to name but a
few. An advanced study must look into this problem and derive analytical and
computational methods to determine the internal moment coefficients for viscous
damping model as a function of the rotor parameters as mentioned. In addition, adding
more body of experimental data on rotor configurations with shrink fit and interference
fit joints can give more insight into the proposed analytical work.
In addition, the computational software such as XLTRCTM should be modified to
include the Coulomb friction model to analyze the stability of the rotor-bearing systems
with shrink fit and interference fit interfaces.
118
REFERENCES
1. Walton, J., Artilles, A., Lund, J., Dill, J., Zorzi, E., 1990, “Internal Rotor Friction
Instability”, Mechanical Technology Incorporated Report, 88TR39
2. Newkirk, B.L., 1924, “Shaft Whipping”, General Electric Review, 27(3), pp.
169-178
3. Jeffcott, H.H., 1919, “The Lateral Vibration of Loaded Shafts in the
Neighbourhood of a Whirling Speed: The Effect of Want of Balance,”
Philosophical Magazine, Ser.6, 37, pp.304
4. Kimball, A.L. Jr., 1924, “Internal Friction Theory of Shaft Whirling”, General
Electric Review, 27(4), pp.244-251
5. Gunter, E.J., 1966, “Dynamic Stability of Rotor-Bearing Systems”, NASA
Technical Report, SP-113
6. Walton, J.F.Jr.,Martin, M.R., 1993, “Internal Rotor Friction Induced Instability in
High-speed Rotating Machinery”, Vibration of Rotating Systems, DE-Vol.60, pp.
297-305
7. Kimball, A.L. Jr., 1925, “Measurement of Internal Friction in a Revolving
Deflected Shaft,”, General Electric Review, 28, pp.554-558
8. Kimball, A.L. Jr., Lovell, D.E., 1926, “ Internal Friction in Solids,” Transactions
of ASME, 48, pp.479-500
9. Lund, J.W., 1986, “Destabilization of Rotors from Friction in Internal Joints with
Micro-slip,” International Conference in Rotordynamics, JSME, pp. 487-491
10. Artilles, Antonio F., 1991, “ The Effects of Friction in Axial Splines on Rotor
System Stability,” IGTA Congress and Exposition, pp. 1-7
11. Black, H.F., 1976, “The Stabilizing Capacity of Bearings for Flexible Rotors
with Hysteresis”, Transactions of the ASME, pp.87-91
12. Ehrich, F.F., 1964, “Shaft Whirl Induced by Rotor Internal Damping”, Journal of
Applied Mechanics, 31, pp. 279-282
119
13. Yamamoto,T., Ishida, Y., Linear and Nonlinear Rotordynamics, 2001, John
Wiley Publications, NY
14. Vance, J.M., Ying, D., “Effects of Interference Fits on Threshold Speeds of
Rotordynamic Instability”, Paper No. 2007, Proceedings of the International
Symposium on Stability Control of Rotating Machinery, August 20-24, 2001,
South Lake, Tahoe, California
15. Mir, Mohammad M., “Effects of Shrink Fits on Threshold Speeds of
Rotordynamic Instability”, MS thesis, 2001, Texas A&M University, College
Station
16. Srinivasan, A., “Effects of Shrink fits on Threshold Speeds of Rotordynamic
Instability”, MS thesis, 2003, Texas A&M University, College Station
17. Murphy, B.T., “Eigenvalues of Rotating Machinery”, PhD dissertation, 1984,
Texas A&M University, College Station
18. Robertson, D., 1935, “Hysteretic Influences on the Whirling of Rotors”,
Proceedings of the Institute of Mechanical Engineers, London, 131, pp. 513-537
19. Smalley, A.J., Pantermuehl, P.J., Hollingsworth, J.R., Camatti, M., 2002, “ How
Interference Fits Stiffen the Flexible Rotors of Centrifugal Compressors,”
IFToMM Sixth International Conference on Rotordynamics
20. Nelson, H.D., McVaugh, J.M., 1976, “The Dynamics of Rotor-Bearing Systems
Using Finite Elements,” Transactions of ASME, Journal of Engineering for
Industry, 98(2), pp. 593-600
21. Zorzi, E.S., Nelson, H.D., 1977 “Finite Element Simulation of Rotor-Bearing
Systems With Internal Damping”, Journal of Engineering for Power, 99, pp. 71-
76
22. Hashish, E., Sankar, T.S., 1984, “Finite Element and Modal Analyses of Rotor-
Bearing Systems Under Stochastic Loading Conditions”, Transactions of the
ASME, Journal of Vibrations, Acoustics, Stress, and Reliability in Design, 106,
pp. 80-89
120
23. Ginsberg, Jerry H., 2001, Mechanical and Structural Vibrations, John Wiley
Publications, NY
24. Meriam, L.G., Kraige, J.C., 1995, Statics, John Wiley Publications, NY
121
APPENDIX A
ANALYSIS OF SHRINK FIT INTERFACE STRESSES FOR ROTATING
CIRCULAR DISKS
Determination of machine stresses at shrink fit interfaces due to shrink fits
constitutes an important study from two points of views: machine design and
rotordynamic instability. From machine design point of view of machines with shrink
fits, it is often required that the components in contact retain their elastic behavior and
are not stressed beyond their plastic limit. Moreover, when such machines execute
rotational motion, it is desired that shrink fits be tight enough, so that the components do
not loose contact with each other due to centripetal effects and resulting looseness of the
contact. From rotordynamic instability point of view, the determination of machine
stresses is important because it enables an engineer to estimate magnitude of the
frictional moments that are generated as micro-slip takes place between shrink fitted
components. It has been experimentally verified [4, 5] that this micro-slip results in
potentially dangerous forward whirl instability of the shrink fitted rotors when they are
operated above their first critical speeds. Therefore, any rotordynamic predictions for
assessing stability of the rotors, especially in regards to predicting threshold speeds of
instability, which require estimate of these contact stresses to formulate frictional
stresses and equivalent frictional moments, are extremely important to ensure safe and
stable operation of machines with shrink fit interfaces. This study is directed at deriving
analytical formulae for an example three-disk model with shrink fits at the two
interfaces. The stresses are assumed to be functions of radial dimension of the materials
involved only (one-dimensional). The equations thus derived show how interface
stresses and shrink fits are functions of component material properties, interface size,
initial shrink fit values and rotational speed.
122
DESCRIPTION OF THE MODEL
The model, shown in Fig.A1, is comprised of a hollow elastic disk (labeled ‘1’),
mounted over another hollow elastic disk (labeled ‘2’), which in turn is mounted over a
solid elastic disk (labeled ‘3’). The thickness of the disks is assumed to be small
compared to their diameters and therefore stresses are assumed to be independent of
axial and circumferential coordinates. An analysis using axisymmetric deformation of
the disks by utilizing basic equations of elasticity has been carried out for this system of
three synchronously spinning bodies, shrink fitted together, each one of which is elastic
and thus experiences radial and tangential deformations due to centripetal forces and
shrink fits. The radii indicated for the three disks in Fig.1 are the pre-assembly radii and
therefore, the shrink fits at the interfaces (at assembly) are given by differences in radii
for the mating disks. In this analysis, these initial shrink fits will be treated as given
(known) parameters along with other material and geometric parameters of the system,
such as Young’s elastic modulii, Poisson’s ratio, and initial radii (a, b and c) of all the
disks. The model can be extended to cover multiple shrink fitted disks.
The importance of this model is in its direct application to the case of the two-
disk rotor used in the experimental research of shrink fits as described in the main body
of dissertation.
123
Fig.A1 Three cylindrical shafts shrink fitted together. Top two are hollow, while the smallest one is solid
DERIVATION OF EQUATIONS FOR STRESSES
An analysis based on equations of elasticity for axisymmetric deformations can
be developed by considering the three shaft model as shown in Fig.1. The analysis
utilizes the principle of superposition, since the assumptions about material and its
deformation are strictly of linearity. For this purpose, consider cylinders ‘1’ and ‘2’ as
shown in Fig.A1, separately, as shown in Fig.A2 below:
1
2
3
a
bc
124
Fig.A2 Cylinders ‘1’ and ‘2’ shown separated (free-body diagrams) subjected to internal shrink fit stresses
STATIC INTERFACE STRESSES
When the three-disk assembly is not rotating and is in static equilibrium, the
interface stresses can be formulated by considering the deformations at the interfaces as
well as magnitudes of the shrink fits and relating them to the radial interface stresses.
From Fig.A1, disk 1 is subjected to the inner stress only (at the interface with
disk 2), whereas disk 2 is subjected to both the inner and the outer stresses. These
stresses are unknowns and need to be determined from system’s given parameters.
The radial displacements in the disks can be related to the initial shrink fit values
through the following equations:
b
c
2
b
a
1
r
125
021 )br(u)br(u δ==−= (A1)
132 )ar(u)ar(u δ==−= (A2)
In terms of the as yet unknown radial stresses at the interfaces, namely ‘P12’ and
‘P23’, the equations for radial displacements can be written as:
)bccb(
EbP
)br(u 122
22
1
121 ν+
−
+== (A3)
)ab(b
)PP(baE
1ab
PbPab
E1
)br(u22
122322
2
222
122
232
2
22
−
−++
−
−−==
νν (A4)
)ab(a)PP(ba
E1
abPbPa
aE
1)ar(u
221223
22
2
222
122
232
2
22
−
−++
−
−−==
νν (A5)
)1(EaP
)ar(u 33
233 ν−
−== (A6)
Substituting equations (A2) through (A6) into equations (A1) and (A2), a system
of two linear equations in the interface stresses ‘P12’ and ‘P23’ is obtained, which can be
expressed as follows:
0232121 PCPC δ=+ (A7)
1234123 PCPC δ=+ (A8)
The coefficients C1, C2, C3, and C4 in equations (A7) and (A8) are defined as
follows:
126
2 2 3 2
2 21 12 2 2 2 2 2
1 2 2
1 1( )b b c b a bCE c b E b a E b a
ν νν − ++= + + +
− − − (A9)
2
222
2
2
222
2
2 E1
abba
E1
abbaC
νν +
−−
−
−
−= (A10)
)ab(a
baE
1E
1ababC
22
22
2
2
2
222
2
3−
+−
−
−
−=
νν (A11)
3
3
2
222
2
2
222
3
4
E)1(a
E1
abab
E1
abaC
ν
νν
−+
+
−+
−
−=
(A12)
The solution of equations (A7) and (A8) can therefore be expressed as follows:
2341
214012 CCCC
CCP
−−
=δδ (A13)
1432
113023 CCCC
CCP
−−
=δδ (A14)
Equations (A13) and (A14) express the values of the static radial interface
stresses as functions of such system variables as magnitude of the initial shrink fits at the
interfaces, and geometric and elastic properties of the disks.
127
(b) Interface radius
The interface radius for the static assembly can be determined by using equations
(A3) and (A5) or equations (A4) and (A6). The interface radii for the disks ‘1’ and ‘2’
after the assembly are:
b)br(uR 112m +== (A15)
a)ar(uR 323m +== (A16)
In equations (A15) and (A16), the radial displacements u1 and u3 can be
calculated using equations (A3), (A6), (A13), and (A14). Therefore, the interface radius
can be computed as a function of the initial shrink fit and the other system parameters.
DYNAMIC STRESSES AND SHRINK FIT Cylinder 1
Let the elastic modulus and Poisson’s ratio of cylinders ‘1’ and ‘2’ be E1, ν1 and
E2, ν2, respectively. From theory of elasticity, the stress distribution equations for
cylinder ‘1’ which is subjected to centripetal stresses as well are given by:
221
121
11r r8
3rB
A)r( ωρν
σ+
−−= (A17)
221
121
11t r831
rB
A)r( ωρν
σ+
−+= (A18)
In equations (A1) and (A2), ‘r’ is the radial coordinate to locate any point
between and including inside and outside radii of the cylinder ‘1’ and ‘ρ1’ is the mass
density of material which comprises cylinder ‘1’. Subscripts ‘r’ and‘t’ denote radial and
tangential stresses, respectively. Also, let ‘a’, ‘b’ and ‘c’ be the radii of the cylinders in
128
their undeformed positions (i.e., before they are shrink fitted together). ‘A1’ and ‘B1’ are
constants to be evaluated from boundary conditions of the problem. The boundary
conditions are as follows:
(1) At r = b, σr1 = -p12 (interface stress due to shrink fit)
(2) At r = c, σr1 = 0 (no external load)
Using these boundary conditions in equation (A1), following equations result:
221
121
112 b8
3bB
Ap ωρν+
−−=− (A19)
221
121
1 c8
3cB
A0 ωρν+
−−= (A20)
From equations (A3) and (A4):
)cb(8
3cB
bB
p 2221
121
21
12 −+
−+−
=− ωρν (A21)
Solving equation (A5) for constant ‘B1’:
2221
11222
22
1 cb8
3p
cbcbB ωρ
ν++
−
−= (A22)
From equation (A4), ‘A1’ can be expressed in terms of ‘B1’as follows:
221
121
1 c8
3cB
A ωρν+
+= (A23)
From equations (A2) and (A7):
129
221
12122
11
21
1t r831
rB
c8
3cB
)r( ωρν
ωρν
σ+
−++
+= (A24)
By utilizing boundary conditions of the problem and equations (A8) and (A6),
the following values for radial and tangential stresses are obtained at the interface radius
‘b’:
221
1221
11222
22
1t b)8
1(2c)
83
(2pbccb)br( ωρ
νωρ
νσ
++
++
−
+== (A25)
121r p)br( −==σ (A26)
Therefore, the radial deformation of the cylinder ‘1’ at interface radius ‘b’ can be
calculated using the modified Hooke’s Law as expressed in polar coordinates as follows:
)br(b)br(u 1t1 === ε (A27)
The tangential strain ‘εt1’ can be expressed as follows:
))br()br((E1)br( 1r11t1
1t =−=== σνσε (A28)
Using equations (A9) and (A10) and substituting them in equation (A12) and
(A11), the following expression for the radial deformation at the interface radius is
obtained:
321
1
1221
1
122
22
11
121 b)
E81
(2bc)E8
3(2]
bccb[
Ebp
)br(u ωρν
ωρν
ν+
++
+−
++== (A29)
130
Cylinder 2
In a similar manner to cylinder ‘1’, equations of stress distribution for inside
cylinder ‘2’ can be written as follows:
222
222
22r r8
3rB
A)r( ωρν
σ+
−−= (A30)
222
222
22t r831
rB
A)r( ωρν
σ+
−+= (A31)
However, the boundary conditions are different in this case because the inner
cylinder is subjected to both internal and external stresses due to shrink fits at both
locations. These can be expressed as follows:
(3) At r = a, σr2 = -p23
(4) At r = b, σr2 = -p12
Using boundary conditions (3) and (4) above, equation (14) yields the following
two equations:
222
222
223 a8
3aB
Ap ωρν+
−−=− (A32)
222
222
212 b8
3bB
Ap ωρν+
−−=− (A33)
Using equation (A16) to express ‘A2’ in terms of ‘B2’ and substituting the result
in equation (A15) and finally calculating the tangential stress at interface radius r = b,
yields the following result:
222
2222
2231222
22
232t b)8
1(2a)
83
(2)pp(babap)br( ωρ
νωρ
νσ
++
++−
−
++−== (A34)
122r p)br( −==σ (A35)
131
Using equations (A18) and (A19), the tangential strain in cylinder ‘2’ at interface
radius ‘b’ is expressed as follows:
))br()br((E1)br( 2r22t2
2t =−=== σνσε (A36)
Therefore, the final result for radial deformation in cylinder ‘2’ (using equations
(A18), (A19) and (A20)) is as follows:
322
2
2222
2
222
22
2
2322
22
22
122 b)
E81
(2ba)E8
3(2]
baba1[
Ebp
]baba[
Ebp
)br(u ωρν
ωρν
ν+
++
+−
++−
−
++==
(A37)
Subtracting equation (A21) from equation (A13) yields the equation for variable
shrink fit as a function of rotational speed ‘ω’ as follows:
321
1
1
322
2
2221
1
1222
2
2021
b)E8
1(2
b)E8
1(2bc)
E83
(2ba)E8
3(2)br(u)br(u)(
ωρν
ωρν
ωρν
ωρν
δωδ
+−
++
+−
++==−==
(A38)
In equation (A22), ‘δ0’ is the shrink fit at zero speed (initial shrink fit) and it
appears due to the interface stress terms that are originally present in equations (A13)
and (A21). Therefore, equation (A22) provides the variation of shrink fit between
outermost cylinders as a function of rotational speed, given some initial shrink fit at
interface at zero speed.
The dynamic stresses at the interface between disks 1 and 2 can be approximated
by using equations (A13) and (A25) as follows:
132
2341
2140r CCCC
CC)()(
−−
=δωδ
ωσ (A39)
)(bccb)( r22
22
1t ωσωσ−
+= (A40)
NORMALIZATION OF INTERFACE STRESSES
The static radial and tangential stresses acting on the disk 1 at the interface
between disks 1 and 2 as shown in equations (A13) and (A25) (by setting ω=0) can be
normalized with respect to the yield stress of the material which comprises disk 1. The
reason for normalizing the stresses is to find out what value of the interface shrink fit δ0
causes the initiation of failure of the material comprising disk 1. This is in accordance
with the maximum shear stress theory as a failure criterion of engineering materials. The
maximum shearing stress criterion states that for ductile materials such as various grades
of steel and aluminum, the failure initiates when the stress acting on the body is equal to
or greater than the yield strength of the material, provided the stresses are of the same
numerical sign, as in the case of outer disk in the model considered. As long as the
stresses are less than the yield stress, the material’s failure will not initiate due to static
loading.
Dividing equations (A13) and (A25) by the yield stress of the material, the
following equations for the normalized stresses are obtained:
)CCCCCC
(1
2341
2140
YY
rr −
−==
δδσσ
σΣ (A41)
22
22
2341
2140
YY
tt
bccb)
CCCCCC
(1−
+−−
==δδ
σσσ
Σ (A42)
133
The graphs of equations (A41) and (A42) are shown on the following pages. On
vertical axis are the normalized stresses, whereas on horizontal axis are the values of the
initial shrink fit δ0. The graphs are plotted for the following set of data:
c = 5 in. b= 4.37 in. a = 1.5 in.
E1 = 10 x 106 psi E2 = E3 = 30 x106 psi
ν1 = 0.28 ν2 = ν3 = 0.30
γ1 = 0.1 lb/in3 γ2 = γ3 = 0.28 lb/in3
δ1= 2 mils σY = 35 ksi
ρ = γ /g
The data values shown above are from the experimental rotor (the two disk rotor).
0 5 10 15 200
0.2
0.4
0.6
0.8
1
1.2
Radial shrink fit, Mils
Nor
mal
ized
tang
entia
l stre
ss
Normalized tangential stress vs. shrink fit
Fig. A3 Normalized tangential stress as a function of radial shrink fit at outer interface
134
Fig.A3 shows the normalized tangential stresses acting on disk 1 as a function of
radial shrink fit at the outer interface (δ0). From Fig.A3, the material of outer disk (in
this case, aluminum) will yield (when the value on vertical axis reads 1.0) when the
radial shrink fit is about 16 Mils. This corresponds to a diametral shrink fit of 32 Mils. In
any of the experiments involving the two disk rotor, the initial radial shrink fit value did
not exceed 7 Mils. Therefore, the aluminum sleeve did not undergo plastic deformation
at assembly.
The radial stresses are much smaller in magnitude. The normalized radial stress
as a function of radial shrink fit (δ0) is shown below:
Normalized radial stress vs. shrink fit
0 5 10 15 200
0.05
0.1
Radial shrink fit, Mils
Nor
mal
ized
radi
al st
ress
Fig. A4 Normalized radial stress as a function of radial shrink fit at outer interface
Fig.A4 shows that the radial stresses acting on the disk 1 (aluminum sleeve) are
well within safe limits for radial shrink fits limits.
135
Fig.A5 shows the graphs of equation (A38) with two different initial shrink fit
values as parameter. Fig.A5 shows that the radial shrink fit at the interface is a
decreasing function of rotational speed. For an initial shrink fit of 3.5 Mils as shown by
the solid line in Fig.A5, the value of the radial shrink fit at a speed of 12000 rpm
decreases to less than 1 Mil. For an initial radial shrink fit of 2.5 Mils, the value drops
down to 0 Mils at 12000 rpm. A value of zero radial fit indicates that the outer disk
looses contact with the inner disk. The analysis developed for the rotating disks is also
important from point of view of predicting before finalizing a design whether the design
is safe in so far as contact of the disks during operating speed range is concerned.
0 2000 4000 6000 8000 1 .104 1.2 .104 1.4 .1041
0
1
2
3
4
Initial fit = 3.5 MilsInitial fit = 2.5 Mils
Rotational speed, RPM
Dyn
amic
shrin
k fit
, Mils
Shrink fit at outer interface vs rotational speed
Fig. A5 Radial shrink fit variation with rotational speed.
136
SINGLE SHRINK FIT INTERFACE MODEL
Fig.A6 Two cylindrical shafts shrink fitted together.
The procedure to analyze the interface stresses and shrink fits for a model
comprised of two circular elastic disks is the same as for the three-disk model considered
earlier. The analysis is considerably simplified due to the presence of only one interface
as compared to two interfaces considered in the previous model. The importance of this
model is in its direct application to the case of the single-disk uniform shaft rotor used in
the experimental research of interference fits as described in the main body of
dissertation. The model shown in Fig.6 consists of two circular elastic disks shrink fitted
with a known amount of radial interference fit. The material properties for the disks are
known. The outer disk (labeled disk ‘1’) is a hollow disk, with an outside diameter Ro,
whereas the inside disk is a solid disk (labeled disk ‘2’) with an initial outside diameter
Ri. The dimensions shown in Fig.A6 are pre-assembly dimensions, that is, when the
disks are not connected through shrink fit. When they are assembled or connected
through shrink fit, the dimensions will change due to elasticity of the materials of disks.
1
2
Ro
Ri
137
STATIC INTERFACE STRESS
For the two disks shown, equations for radial displacements at the interface can
be written. The equations are written based on the assumption that the outer disk (disk 1)
is subjected to only inner interface stress (no outer or external stress) whereas the inner
disk (disk 2) is subjected to external stress (which is the interface stress) at its outer
surface.
The radial displacement of the outer disk at the interface is given by:
)RR
RR(
EpR
u 12i
2o
2o
2i
1
12i1 ν+
−
+= (A43)
)1(E
pRu 2
2
12i2 ν−
−= (A44)
In equations (A43) and (A44), p12 is the radial interface pressure. If the disks 1
and 2 are made of same material, the analysis can be simplified. The difference of the
radial displacements is equal to the initial value of the radial shrink fit, that is:
)RR(E
pRR2uu
2i
2o
122
oio21
−==− δ (A45)
Using equation (A45), the static radial interface pressure can be calculated as:
2
oi
2i
2oo
12RR2
)RR(Ep
−=
δ (A46)
The static tangential stress at the interface can be calculated using the radial
stress. From the theory of thick disks, the static radial and tangential stress distributions
as a function of radial coordinate are given as:
138
21
1rrB
A)r( −=σ (A47)
21
1trB
A)r( +=σ (A48)
Using equation (A47), the following boundary conditions can be applied:
(a) At r = Ro, σr (Ro) = 0
(b) At r = Ri, σr (Rm) = -p12
In the boundary condition above, Rm is the interface radius and its value can be
computed from the following equation:
2im uRR += (A49)
The following two equations result from the application of boundary conditions
on equation (A47):
2
o
11
R
BA0 −= (A50)
2
i
11
R
BA12p −=− (A51)
Solving equations (A50) and (A51) simultaneously for the constants A1 and B1:
)RR
RR(pB
2m
2o
2m
2o
121−
= (A52)
139
)RR
R(pA
2m
2o
2m
121−
= (A53)
Using equations (A48), (A52) and (A53), the tangential stress distribution at the
interface can be calculated as follows:
)RR
RR(p)Rr(
2m
2o
2m
2o
12mt−
+==σ (A54)
NORMALIZING THE INTERFACE STRESSES
The radial and tangential stresses given by equations (A46) and (A54) can be
normalized by the yield stress of the material forming disks 1 and 2 to provide
knowledge of what radial shrink fit value causes yield of the material at the interface.
Dividing equations (46) and (54) by the yield stress, the normalized form of the
equations is obtained as follows:
2
oiY
2i
2oo
Y
12r
RR2
)RR(Ep
σ
δσ
Σ−
== (A55)
)RR(
)RR(
RR2
)RR(E2
m2
o
2i
2o
2oiY
2i
2oo
Y
tt
−
−+==
σ
δσσ
Σ (A56)
The graphs of equations (A55) and (A56) are plotted on the next pages. For AISI
4340 steel, the yield strength is about 50 ksi. The outside diameter of the disk is taken as
10 in. (Ro= 5 in.) and the inside diameter is taken as 2.5 in. (Ri = 1.25 in.).
140
0 1 2 3 40
0.2
0.4
0.6
0.8
1
1.2
Shrink fit, Mils
Nor
mal
ized
tang
entia
l stre
ss
Normalized tangential stress
Normalized radial stress
0 1 2 3 40
0.2
0.4
0.6
0.8
Shrink fit, Mils
Nor
mal
ized
tang
entia
l stre
ss
Fig. A7 Normalized tangential and radial interface stresses for disk 1
141
Fig.A7 shows the normalized tangential and radial stresses as functions of radial
shrink fit δo. From Fig.A7, the material yield stress is approached at a radial shrink fit
value of 4 Mils. In the experiments conducted on the single-disk rotor, the maximum
interference fit that was maintained between the shaft and the disk through the tapered
sleeve was 2.2 Mils radial. Therefore, the outer disk did not undergo any plastic
deformations during the experiments.
DYNAMIC SHRINK FIT
The procedure to formulate the shrink fit as a function of rotational speed is the
same as that used in the three-disk model. The formulation begins with equations (A17)
and (A18). Appropriate boundary conditions are applied using equation (A17) and a set
of linear equations is obtained in the constants A1 and B1. Then the displacements are
formulated in terms of those constants. The difference in the displacements at a speed
give an expression for the variation of shrink fit with the rotational speed. For the model
of two elastic disks, the expression for dynamic shrink is obtained using this procedure
as follows:
2oi
21o RR)
E83
(2)( ρων
δωδ+
−= (A57)
The graph of equation (A57) is plotted on the next page for three different values
of the initial radial shrink fit. The material is steel. The inside radius is 1.25 in. The
outside radius is 5 in.
142
0 2000 4000 6000 8000 1 .10 4 1.2 .10 4 1.4 .10 41
1.5
2
2.5
3
3.5
4
Initial fit = 2 MilsInitial fit = 3 MilsInitial fit = 4 Mils
Speed, RPM
Shrin
k fit
, Mils
Dynamic Shrink Fit
Fig. A8 Variation of radial shrink fit as a function of rotational speed for the case of
two-disk model
Fig.A8 shows the shrink fit at the interface between disks 1 and 2 as a function of
rotational speed for three different values of the initial shrink fit values. Fig.A8 shows
that the radial shrink fit is a decreasing function of rotational speed. The higher the
initial value of the shrink fit, the higher its value will be at different rotational speeds as
compared to the one with the lower initial values.
143
APPENDIX B
FREE-FREE TESTS ON SINGLE-DISK ROTOR WITH THE DISK AT
VARIOUS AXIAL LOCATIONS ALONG THE SHAFT
Fig.B1 Free-free testing of the single-disk rotor.
Free-free tests of the single-disk rotor were performed to estimate the amount of
internal friction in the rotor. In each of the experiments, the axial location of the disk
was varied along the span of the shaft, whereas the interference fit was maintained the
same at each axial location. The objective of these experiments is to ascertain whether
the internal damping coefficient varies with the axial location of the disk on the shaft,
when the interference fit is maintained the same. For each of the axial locations of the
disk on the shaft, the free-free time data of the rotor vibration was obtained.
144
The free-free vibration data of the single-disk rotor when the disk is at the centre
and when it is near the end of the shaft is shown in Fig.B2 as follows:
Acceleration vs time (Disk at the Centre)
-10
-5
0
5
10
-0.01 0.01 0.03 0.05 0.07 0.09 0.11 0.13 0.15
Time (second)
Acce
lera
tion
(V)
Acceleration vs Time (Disk 15 inches off the centre)
-5-4-3-2-10123456
-0.01 0.01 0.03 0.05 0.07 0.09 0.11 0.13 0.15
Time (seconds)
Acc
eler
atio
n (E
U)
Fig.B2 Free vibration data for the single-disk rotor with different positions of the
disk
145
Fig.B2 shows that changing the axial position of the disk considerably affects the
amount of friction which is developed in the rotor due to slippage in the interference fit
interface. There is considerably more amount of friction which is developed for the disk
offset from the centre than in the case when the disk is at the centre, while maintaining
the same interference fit in each of the case.
Through the experimental free-free vibration data, the logarithmic decrement can
be calculated which in turn yields the value of the internal damping coefficient
corresponding to each of the axial locations of the disk on the shaft. The value of the
interference fit that was maintained constant through the experiments was 1.16 mils,
radial. This value of the interference fit corresponds with the initial value of the fit that
caused the rotor-bearing system to become unstable at 5800 rpm, when the rotor was
mounted on ball bearings.
The data for the free-free experiments was collected using a dynamic signal
analyzer and an accelerometer. The accelerometer was attached at the centre of the disk
in the horizontal direction and the rotor was tapped through a soft hammer. In each of
the experiments, the time traces of acceleration as measured from the accelerometer
were obtained for the first mode of free vibration of the rotor. To ensure that only the
data for the first mode is obtained, the signal analyzer was also setup to display the
frequency spectrum on linear scale. In a frequency spectrum, the amplitude
corresponding to each frequency is displayed. From the frequency spectrum, it was made
sure that a high vibration amplitude of only the first mode of free-free vibration of the
rotor was obtained (by trial and error, through tapping the rotor at different locations
until only the first mode was visible on the spectrum plot) and that the other modes were
either not excited, or else the vibration amplitudes corresponding to the higher modes
were negligible as compared to the first mode. In this way, the logarithmic decrements
and the damping associated with only the first modes for each of the experiments was
obtained.
The experimental results from the single-disk rotor are summarized in Table 1 on
the next page:
146
Table B1.Experimental results from free-free tests of the single-disk rotor.
Table 1 shows that the first mode frequency of the single-disk rotor increases as
the disk is made progressively offset from the centre of the shaft. As the disk is moved
towards the coupling end side, the average logarithmic decrement and the corresponding
damping ratios decrease. On the other hand, if the disk is moved towards the coupling
end of the shaft away from the centre, the logarithmic decrement and the corresponding
damping ratio increases. This opposing trend can be explained by the interface geometry
of the tapered sleeve and the shaft. The tapered sleeve is more likely to slip on the shaft
at one end as compared to the other end because it is fastened to the wheel through the
draw bolts. As the end of the sleeve which is opposite to the fastening end moves away
from the centre, it experiences more micro-slip at the interface due to the mode shape of
the rotor. At the ends of the rotor farther away from the mid-point of the shaft, the rotor
has higher amplitudes in its first mode, as it has high amplitude at the mid-point. In
between the mid-point and the farther ends of the shaft, there is a node located. When
the sliding end of the tapered sleeve is at the farther ends of the shaft, higher micro-slip
takes place and as a result, the damping is higher. On the other end, when the non-
fastening end of the tapered sleeve (inside the wheel) is at a node location in the first
Disk Location First mode frequency (Hz) Logarithmic Decrement Damping Ratio
Centre 128 0.1004 0.016
4 in. left 136 0.087 0.0138
8 in. left 148 0.054 0.0086
12 in. left 152 0.0866 0.014
5 in. right 132 0.068 0.0109
11 in. right 148 0.259 0.041
15 in. right 152 0.129 0.0205
147
mode, the micro-slip is negligible and therefore, the damping is small as compared to
when the disk is either at the centre or at the extreme ends of the shaft.
The average moment damping coefficients are calculated from the average
damping ratios as presented in Table 1 using the following equation:
[ ] [ ]1RotT
111 C2 ΦΦως = (1)
In equation (1), the matrices on the right hand side are the normal mode matrices
of the rotor in its first mode of free-vibration. The variables ζ1 and ω1 are the damping
ratios and the first mode natural frequency respectively which are listed in Table 1. The
frequency given in Table 1 is in Hz, whereas the frequency used in equation (1) is in
rad/s. The frequency ω1 can be obtained from the measured frequency by multiplying it
with the factor 2π. The computation of the normal mode matrices is performed using the
XLTRCTM software. The models of each of the different single-disk rotor configurations
are simulated as free-free rotors and their mode shapes are computed using the “Shapes”
feature in the XLTRCTM software. The result of using equation (1) to compute the
moment damping coefficients for various configurations is summarized in Table B2:
148
Table B2. Calculated values of the moment damping coefficient at various disk locations
Disk Location First Mode (Hz) Moment damping(lb-in-s)
Centre 128 124.22
4 in. left 136 359.34
8 in. left 148 613.51
12 in. left 152 54.42
5 in. right 132 68.83
11 in. right 148 1678.45
15 in. right 152 841.58
149
APPENDIX C
XLTRCTM ROTORDYNAMIC SIMULATIONS OF IMAGINED ROTOR
CONFIGURATIONS
This Appendix describes the rotordynamic simulations of several imagined rotor
configurations (both single-disk and two-disk rotors) with shrink fit joints. The objective
of presenting these simulations is an attempt to understand the effect of various system
parameters such as geometry and material properties on the stability of the system. As
shown in Appendix B, the location of disk at various axial locations of the shaft results
in different modal damping. The simulations of various imagined configurations of
single-disk and two-disk rotor with varying geometries and material properties using the
internal moment model to simulate shrink fit joint internal friction are presented and
discussed. None of these simulations (except one two-disk rotor model), however, is
supported by any experimental data. The values of the internal moment coefficients,
such as the direct moment stiffness and direct moment damping coefficients, used to
simulate internal friction in each of the cases, are based on engineering judgment and
should not be considered as “final” values to simulate the imagined rotor configurations.
Indeed, the word “imagined” is supposed to imply the fact that there is no experimental
data on rotordynamic stability to support these simulations which are presented and
discussed in this Appendix. The simulations in this Appendix merely serve as a
guideline for designing a single-disk and a two-disk rotor with shrink fit interfaces and
to assess the rotordynamic stability trend in similar configurations of rotating machines.
150
SINGLE-DISK IMAGINED ROTOR CONFIGURATIONS AND SIMULATIONS
1. Disk located 1/3rd the length of shaft from the coupling end
This is the first of imagined rotor configuration simulations presented. In this
configuration, all the dimensions and elastic properties of the experimental setup are
retained except for the axial position of the disk on the shaft. As shown in Appendix B,
the axial location of the disk on shaft results in different damping ratios and
consequently, different modal damping. The ‘Geo Plot’ of the rotor model from
XLTRCTM is shown in Fig. C1:
Fig.C1 Single-disk rotor model for disk offset from the centre of the shaft
151
The simulation of the rotor model shown in Fig.C1 shows that the model is more
unstable as compared to the baseline model as shown in Chapter VI. The first critical
speed of the rotor is 3590 rpm. The first critical speed of the baseline case as measured
experimentally and predicted theoretically (Chapter II and VI) is around 3000 rpm. The
increase in the first critical speed of the imagined rotor model is due to stiffening as
introduced by shifting the disk to a location different from the centre of the shaft.
However, the model is more unstable as a smaller amount of direct moment damping is
required to predict a threshold speed of instability as low as 3800 rpm, which is just
above the predicted critical speed. The value of the direct moment damping coefficient
which predicts this threshold speed of instability is 6207 lb-in-s. A comparison with the
direct damping value for this rotor model and the one from Table 1 shows that the
smaller direct damping value in the disk-offset rotor model still results in a substantially
more unstable rotor as compared to the one at the disk centre.
The higher unstable behavior of the disk-offset from centre model can be
explained by considering the first mode shape of the rotor on the bearings. At the centre,
the slope of the mode shape is zero whereas at a position offset from the centre, the slope
of the mode shape is non-zero. When the friction interface is closer to the centre, the
difference in the rotational deflection of the disk and the shaft will be smaller as
compared to when the disk is offset. This is because when the rotor bends in the first
mode shape, the disk tends to maintain its perpendicularity with respect to deflected
centerline of the shaft. At a location near the centre, the slip will be very small due to
disk maintaining its perpendicularity. As a result, the magnitude of the resulting internal
moments will be smaller as compared to when the disk is offset from the centre, due to
greater amount of relative slip between the disk and the shaft. Since internal moment
will be larger in the second case, it will make this system more unstable as compared to
the baseline case.
152
2. Disk very close to bearing
Fig.C2 Single-disk rotor model for disk very close to a bearing location
The rotor model with the disk offset from the centre of the shaft and located very
close to a ball bearing location is shown schematically in Fig.C2. The remaining shaft
and disk properties are the same as for the baseline case of the single-disk rotor model as
discussed in Chapter VI. The XLTRCTM simulations of the rotor model shown in Fig.C2
show that this model is more unstable as compared to the baseline model (with disk at
the centre). That is, using a smaller direct moment damping coefficient (around 6500 lb-
in-s) yields the threshold speed of instability which is slightly smaller than the threshold
speed of the baseline model (5800 rpm). The simulated first critical speed of the model
in Fig.C2 is about 4200 rpm. The increase in the first critical speed is an indication of
153
stiffening of the rotor-bearing system in Fig.C2 as compared to the baseline case (for
which the first critical speed is 3000 rpm). The increase in the first critical speed can be
explained by the effect of moving the disk away from the centre. When the disk is not at
the centre, the modal mass involved in the first mode of vibration is decreased and
therefore the first critical speed is increased. The higher instability can be explained by
the free-free experimental testing of the single-disk rotor as presented in Appendix B.
The farther is the friction producing interface from the centre, the more slipping will
occur at the sleeve-shaft interface, resulting in large internal moments and accordingly
higher instability.
3. Disk at mid-span, axial thickness of disk = 2.5 inches
Fig.C3 Single-disk rotor model with axial width of disk = 2.5 inches
154
The single-disk rotor model with reduced axial width of the disk (2.5 inches, as
compared to 5 inches for the baseline case) and higher outside diameter to keep the
weight of the disk the same as the baseline case. The XLTRCTM simulations of the rotor
model show the first critical speed to be 3000 rpm, which is the same as for the baseline
case. This is to be expected, because the weight of the disk as well its axial location
relative to the shaft is unchanged. However, the rotor model is more stable as compared
to the baseline case. The model requires a higher direct moment damping coefficient to
predict the same threshold speed of instability as compared to the baseline case. This can
be explained in view of experiments conducted on the single-disk rotor as described in
Appendix B. The closer the friction interface is to the shaft centre, the lesser is the
amount of modal damping. Therefore, the rotor will be more stable in this configuration,
with a smaller direct moment damping coefficient as compared to the baseline case.
155
4. Disk at mid-span, axial thickness of disk = 1 inch
Fig.C4 Single-disk rotor model with axial width of disk = 1 inch
The single-disk rotor model with the disk axial width reduced to 1 inch (as
compared to the baseline case of 5 inches) with the same weight of the disk as the
baseline case is shown in Fig.C4. The simulations of the rotor model using the
XLTRCTM software show that the first critical speed of the rotor is around 2700 rpm,
which is close to the first critical speed of the baseline case. The slight difference in the
first critical speed of the baseline model and the model in Fig.C4 arises due to some
stiffening effect of the baseline model due to its larger axial width (5 inches).Due to its
larger axial span, the disk in the baseline model increase the effective shaft stiffness. The
model in Fig.C4 can be called as an extended Jeffcott rotor model. The XLTRCTM
simulations show that this rotor model is more stable as compared to the baseline single-
156
disk model, with a higher amount of direct moment damping coefficient required to
predict the same threshold speed of instability as the baseline case.
5. Disk at mid-span, axial width of disk = 10 inches
Fig.C5 Single-disk rotor model with disk axial width = 10 inches
The single-disk rotor model with the disk at mid-span of the shaft and axial width
increased to 10 inches (instead of the baseline case of 5 inches) with the same weight as
the baseline model disk is shown in Fig.C5. The XLTRCTM simulations of the rotor
model show that the first critical speed is 3600 rpm, as compared to the baseline case of
157
3000 rpm. The increase in the first critical speed can be attributed to the increase in
stiffness of the rotor due to axial width of the disk. The disk’s bending stiffness
contributes to the overall increase of stiffness of the rotor. Therefore, even though the
weight of the disk and the shaft are the same as the baseline case, the bending stiffness
of the disk contributes to the overall stiffening of the rotor. The XLTRCTM simulations
show that the model shown in Fig.C5 is more unstable as compared to the baseline
model. This can again be explained in view of the free-free experiments on the single-
disk rotor as explained in Appendix B. Due to larger axial span of the disk, the friction
interface is farther away from the centre of the shaft, resulting in larger slip and larger
friction moments. Even though increasing the axial width of the disk stiffens the rotor, it
also results in larger friction in the system, resulting in a more unstable system.
158
6. Disk at mid-span, axial width of disk = 15 inches
Fig.C6 Single-disk rotor model with disk axial width = 15 inches
The single-disk rotor model with axial width of the disk increased to 15 inches,
while having the same weight as the baseline rotor model, is shown in Fig.C6. The disk
is in this model is more of a sleeve than a disk due to its larger axial span. It can be
expected that the increased axial width will increase the stiffness of the system, resulting
in larger first critical speed. This conclusion is verified by simulations using XLTRCTM.
The simulations show that the first critical speed of the rotor is 4200 rpm, as compared
to 3000 rpm for the baseline case. The increase in the stiffness of the rotor is due to the
contribution of bending stiffness of the disk, which behaves more like a sleeve in this
159
model. In the baseline rotor case, since the axial span of the disk is smaller (5 inches as
compared to 15 inches for this rotor model), it affects the mode shape of the rotor by
forcing it to remain more flat for the portion of the shaft inside the disk, thereby
stiffening the rotor and consequently, increasing the first critical speed.
Although the stiffness of the rotor system is increased, the system is more
unstable as compared to the baseline case. This can again be explained in light of the
experiments presented in Appendix B. Since the friction interface is further away from
the centre of the shaft, more slipping will take place as compared to when it is near the
centre of the shaft. As a result, the rotor will be more unstable for larger axial width of
the disk.
160
7. Disk at mid-span; disk made of aluminum
Fig.C7 Single-disk rotor model with disk at mid-span and aluminum as disk material
The single-disk rotor model with the disk at mid-span of the shaft is shown in
Fig.C7. This rotor is the same as the baseline model except that the material of the disk
is aluminum instead of steel. It can be expected that the rotor-bearing system will have a
higher natural frequency and a higher first critical speed as compared to the baseline
case due to decreased modal mass of the system, which mainly comes from the disk in
the baseline case. The modal mass is reduced in the model shown in Fig.C7 because the
disk is made of aluminum which has a weight density almost 33% of the weight density
of steel. The simulations using XLTRCTM software verify this conclusion. The predicted
161
first critical speed of the rotor model is around 4200 rpm as compared to 3000 rpm for
the baseline case. The simulations show further that the system is marginally more stable
as compared to the baseline case. A higher amount of direct moment damping
coefficient as compared to the baseline case is required to predict the threshold speed of
instability around 5800 rpm, which is the threshold speed of instability for the baseline
case. The increased stability can be explained by considering the increase in the first
critical speed of the rotor model. Even though the first critical speed is increased due to
reduced modal mass of the rotor, the bending stiffness of the rotor is not increased as
compared to the baseline model. Furthermore, the friction interface is still close to the
centre of the shaft as in the baseline case. These factors combine to make the rotor model
in Fig.C7 more stable as compared to the baseline case.
162
8. Disk at mid-span; disk and shaft made of aluminum
Fig.C8 Single-disk rotor with disk at the mid-span; disk and shaft made of aluminum
The single-disk rotor model with disk and rotor shaft both made of aluminum is
shown in Fig.C8. The other geometric and foundation properties are the same as in the
baseline case. The XLTRCTM simulations of the rotor model shown in Fig.C8 show that
the first critical speed of the rotor is 3000 rpm. The rotor model is more unstable as
compared to the baseline case as well as the case 7 discussed. This can be explained
through the material elastic properties. Since the material of the shaft and the disk is
made of aluminum, the modulus of elasticity of the rotor is about 33% of the modulus of
elasticity of steel. Even though the modal mass of the rotor is decreased due to reduced
163
weight of the shaft and the disk, the stiffness of the system is also decreased because
aluminum is nearly 3 times more flexible as compared to steel. As a result, the effects of
decrease in stiffness and the decrease in mass cancel each other out, and the resulting
first critical speed of the rotor is the same as for the baseline case, which is 3000 rpm.
The higher instability of the rotor model can be explained by considering the shaft
material. The aluminum shaft in model of Fig.C8 will bend more easily under the
influence of internal friction moments as compared to the steel shaft for the baseline
case. This is due to lower elastic modulus of the aluminum material. Therefore, the
system will be more unstable as compared to the baseline case.
9. Disk at mid-span; shaft made of aluminum
Fig.C9 Single-disk rotor model with steel disk at mid-span and shaft made of aluminum
164
The single-disk rotor model with the disk at the mid-span of the shaft is shown in
Fig.C9. In this rotor model, the material of the disk is steel, whereas the material of the
shaft is aluminum. The XLTRCTM simulations of the rotor model shown in Fig.C9 show
that the first critical speed of the rotor is 2000 rpm, as compared to 3000 rpm for the
baseline case. The simulations also show that the rotor is more unstable as compared to
the baseline case. The simulation results can be explained by considering the material
elastic properties for the shaft and weight of the disk. In the model shown in Fig.C9,
even though the weight of the disk remains the same as for the baseline case, the
stiffness of the rotor is about 33% of the baseline case due to aluminum material for the
shaft. Thus the stiffness is reduced whereas the modal mass remains almost the same.
This explains the decrease in the first critical speed of the rotor. The higher instability of
the rotor model can be explained by considering the flexibility of the shaft. The
aluminum shaft for the rotor model in Fig.C9 will bend more easily under the action of
internal friction moments, causing the system to be more unstable.
165
10. Disk axial width= 15 inches; disk at mid-span, shaft diameter at centre = 1.5
inches
Fig.C10 Single-disk rotor with stepped shaft and disk axial width = 15 inches
The single-disk rotor model with the disk axial width of 15 inches and a stepped
steel shaft is shown in Fig.C10. The diameter of the shaft at the central portion is 1.5
inches, whereas the diameter at the end portions is 4 inches. This geometry of the shaft is
widely different from the geometry of the baseline case. The outside diameter of the disk
is such that the weight of the disk in rotor model shown in Fig.C10 is equal to the
baseline case. The material of the disk and shaft is steel. The XLTRCTM simulations
show that the first critical speed of the rotor is 2000 rpm. The simulations also show that
166
the rotor is more unstable as compared to the baseline case. This is due to a thinner shaft
section in the central portion of the shaft. Since the internal friction moment is applied at
the interface between the disk and the central section of the shaft (between points 13 and
30), the steel shaft bends relatively more under the action of internal moments as
compared to the shaft with a diameter of 2.5 inches. The bending stiffness is
proportional to the fourth power of diameter; reducing the diameter of the central section
from 2.5 inches to 1.5 inches increases the flexibility of the shaft and therefore the rotor
model is more unstable as compared to the baseline case.
11. Disk axial width= 15 inches; disk at mid-span, shaft diameter at centre = 2.5
inches
Fig.C11 Single-disk rotor with stepped shaft and shaft diameter = 2.5 inches at centre
167
The single-disk rotor model shown in Fig.C11 has stepped steel shaft with 4
inches and 2.5 inches diameter. The disk is 15 inches long, with the weight of the disk
the same as in the baseline case. The model in Fig.C11 is similar to the model in
Fig.C10, except that the central shaft diameter is 2.5 inches. The XLTRCTM simulations
show that the first critical speed of the rotor is 4500 rpm. The increase in the first critical
speed is due to stiffening effects of the disk and the 4 inch diameter sections of the steel
shaft. The simulations further show that the rotor is more unstable as compared to the
baseline case. This is due to the larger axial width of the disk, resulting in larger slip at
the friction interface. In addition, between the 4 inch diameter shaft section and the 2.5
inch diameter section there is more difference in slope of the mode shape of the rotor in
its first mode. The two factors combine to produce more slip at the friction interface,
resulting in larger magnitude of the internal moments and correspondingly larger
instability of the rotor.
168
12. Disk axial width= 15 inches; disk at mid-span, shaft diameter = 4 inches
Fig.C12 Single-disk rotor with shaft diameter = 4 inches
The single-disk rotor shown in Fig.C12 has the disk at the mid-span of the shaft.
The weight of the disk is the same as in the baseline model; however, the diameter of the
rotor shaft is increased to 4 inches. The simulations using XLTRCTM show that the rotor
is stable in the speed range of 0-12000 rpm. The simulated first critical speed of the rotor
is 6600 rpm in the vertical direction (Y direction) and 5100 rpm in the horizontal
direction (X direction). The simulation shows that the rotor-bearing system will gain
stability against the destabilizing effects of internal friction due to shrink fits if the
169
diameter of the shaft is increased. This can be explained by considering the action of
internal friction moments on the bending of the shaft. As the shaft diameter is increased,
the shaft bending stiffness is increased and therefore it requires a larger magnitude of
internal friction moments to bend the shaft. If the internal moments are not large enough
in magnitude, the shaft will not bend in the direction of forward whirl above the first
critical speed under the action of internal friction moments, resulting in stability of the
rotor-bearing system.
170
TWO-DISK ROTOR SIMULATIONS
Before describing the imagined two-disk rotor configurations, an experimental
configuration from reference [13] is described and its rotordynamic simulation using the
XLTRCTM software is discussed.
1. Vance-Ying rotor
This rotor was constructed by some modifications made to a Bently rotor kit
[13]. The test rotor consisted of an aluminum sleeve, which was 7.5 inches long with
2.988 inches inside diameter, and 1/8 in. thick. Two identical disks of 2.975 inches
outside diameter and ¾ in. thick were mounted on a thin steel shaft 3/8 in. in diameter.
The shaft was simply supported in two ball bearings (FAFNIR, model S3K).The
distance between two ball bearings supports was 16 inches. A set of orthogonal
proximity probes (X and Y) was mounted close to the left disk to measure the rotor
vibration. A feedback controlled motor was used to run the rotor to speeds above 10,000
rpm.
Fig. C13 Sketch of Vance-Ying rotor with sleeve and shaft dimensions indicated
7.5 in.
16.0 in.
3/8 in. dia.
171
In this rotor, the weight of the individual parts, based on approximate calculation
from the values of dimensions of the parts, can be presented as follows:
(a) Disks: 2.90 lb for both disks
(b) Shaft: 0.50 lb
(c) Sleeve weight: 0.91 lb
As reported in [13], the rotor’s natural frequency (on bearings) was measured before
and after the aluminum sleeve was mounted on the wheels. It was found out that the
rotor’s natural frequency in the X direction did not change much (45 Hz without sleeve
as compared to 46.3 Hz with sleeve), but the natural frequency in the Y direction was
increased by 5.6 Hz when the sleeve was put on (45 Hz without sleeve as compared to
50.6 Hz with sleeve). This indicates that the stiffness of the sleeve has a stronger effect
on the natural frequency than the mass of the sleeve. This can be further explained by
observing the mass of each component of the rotor. Since the sleeve mass is 0.91 lb,
whereas the disks contribute 2.90 lb to the total rotor mass, therefore the disks masses
are the main factor for determining the natural frequency. On the other hand, the sleeve
has some thickness, which adds to the bending stiffness of the rotor. However, at the
same time, this bending stiffness will be offset, to some extent, by almost 1 lb weight
contribution (0.91 lb weight of the sleeve). If the shaft were heavier and made of larger
diameter, than the increase in natural frequency would be higher with addition of the
sleeve, due to increased bending stiffness of the sleeve, but not much substantial
contribution from the weight of the sleeve. The model of Vance-Ying rotor analyzed
using the XLTRCTM software is shown in Fig.C14:
172
Fig.C14 XLTRCTM model of Vance-Ying rotor
Simulation of the rotor shows that with symmetric fits (same internal moments at
both interfaces), the rotor is unstable with substantially lower magnitudes of cross-
coupled internal moments as compared to the case of experimental two-disk or single-
disk rotor. This is most likely due to thin flexible rotor shaft. The more flexible the shaft
is, the more likely it is to bend under the action of smaller internal friction moments
developed at the interfaces. With selected internal moment parameters, the predicted first
critical speed is 3000 rpm and the onset speed of instability is also 3000 rpm, with
instability frequency around 2700 cpm and increasing to 2900 cpm at higher speeds.
173
2. Sleeve outside diameter = 10 inches; central shaft diameter = 3 inches
Fig.C15 Two-disk rotor model with diameter of central portion of shaft = 3 inches
The two-disk rotor model shown in Fig.C15 has the diameter of the central
portion of the shaft as 3 inches, as compared to the experimental case, in which the
diameter of the central portion of the shaft (inside the aluminum sleeve) is 2 inches. The
rotor is modeled as having a tight interference fit at one interface (points 10-34,12-36)
with no cross-coupled moments and an under-cut loose fit with cross-coupled moments
at the other interface (points 22-44, 24-46), just as in the baseline model for the
experimental configuration. The rotordynamic simulations using the XLTRCTM software
174
show that the rotor is stable in the speed range of 0-12000 rpm. The stability is a result
of increase in the shaft diameter of the shaft, that makes the bending stiffness of the rotor
to increase and therefore the internal friction moments are not large enough to bend the
shaft and induce instability in the rotor-bearing system.
3. Sleeve outside diameter = 10 inches; central shaft diameter = 4 inches
Fig.C16 Two-disk rotor model with diameter of central portion of shaft = 4 inches
The model shown in Fig.C16 is similar to that in Fig.C15, except that the shaft
diameter has been further increased to 4 inches in the central portion. The simulations
175
using the XLTRCTM software show that the rotor is completely stable in the speed range
of 0-12000 rpm. This shows that increasing the shaft diameter is one of the most
effective ways to increase dynamic stability of the rotor-bearing systems against the
effects of internal friction.
4. Shaft length = 60 inches; sleeve length same
Fig.C17 Two-disk rotor model with increased shaft length (increased by 8 inches)
The model shown in Fig.C17 has an increased shaft length as compared to the
baseline model. The length of the shaft is 60 inches whereas in the baseline model, the
176
shaft length is 52.5 inches. The distance between the bearings (42 inches) is the same in
the model shown in Fig.C17 as compared to the baseline case. The simulations using the
XLTRCTM software show that the rotor is more unstable as compared to the baseline
case. This can be explained due to increase in the deflection of the shaft caused by the
increase in its length. Under the action of internal friction moments, the bending of the
shaft will increase more, giving rise to larger slip at the friction interface that will result
in instability of the rotor.
5. Sleeve made of steel
Fig.C18 Two-disk rotor model with sleeve made of steel
177
In the two-disk rotor model shown in Fig.C18, the rotor geometry and internal
friction moments distribution is the same as in the baseline model, only the sleeve
material is changed from aluminum to steel. The rotordynamic simulations using the
XLTRCTM software show that the rotor is more stable as compared to the baseline
model, with the threshold speed of instability occurring at 11500 rpm, with the unstable
mode frequency at 6034 cpm. The simulated first critical speed of the rotor is 6160 rpm.
The increased stability of the rotor can be explained in terms of the stiffening effect
caused by steel sleeve as compared to the aluminum sleeve in the baseline model. Steel
is almost three times stiff as compared to aluminum; when the internal friction moments
are applied to the rotor model, the moments required to bend the rotor in the direction of
forward whirl will be higher as compared to the baseline model. This explains the
increased stability of the rotor-bearing system shown in Fig.C18.
178
6. Sleeve length = 22 inches
Fig.C19 Two-disk rotor with sleeve length = 22 inches
Fig.C19 shows a two-disk rotor model with the shaft length the same as in the
baseline model (52.5 inches), but the sleeve length increased to 22 inches as compared to
18 inches in the baseline model. The lengths of other portions of the shaft are
accordingly adjusted to make the total length equal to 52.5 inches. The distribution of
the internal friction moments are the same as in the baseline model. The rotordynamic
simulations using the XLTRCTM software show that the rotor is slightly more unstable as
compared to the baseline case. The increased instability can be explained by the effect of
179
flexibility caused by the central portion of the shaft which is inside the sleeve. The
central portion of the shaft which has a diameter of 2 inches can bend slightly more
under the action of internal friction moments as compared to the baseline model. The
simulated first critical speed of the rotor is 7100 rpm, and the predicted threshold speed
of instability is 8500 rpm with the frequency of unstable mode as 7062 cpm.
7. Sleeve length = 10 inches
Fig.C20 Two-disk rotor with sleeve length = 10 inches
180
Fig.C20 shows a two-disk rotor model with the sleeve length reduced to 10
inches as compared to 18 inches in the baseline model. The total shaft length is the same
as the baseline case (52.5 inches) and the distribution of internal friction moments is also
the same as in the baseline model (tight fit at one interface with no cross-coupling; loose
fit with substantial magnitude of cross-coupling at the other end). The rotordynamic
simulations using the XLTRCTM software show that the threshold speed of instability is
10500 rpm. The first critical speed of the rotor model is 5500 rpm. The stability of the
rotor model in Fig.C20 as compared to the instability of the baseline model can be
explained by considering the stiffening of the rotor due to decrease in length of the
sleeve and central portion of the shaft. Decreasing the length of the central portion of the
shaft (which has a diameter of 2 inches) results in decrease in bending deflection of the
rotor under the action of internal friction moments at the interface. The rotor model is
therefore more stable since the shaft will not bend in the direction of forward whirl under
the action of internal moments. This rotor model shows that the effect of decreasing the
central portion of the shaft is just opposite for the previous rotor model case (case 7, in
which the sleeve length is increased from 18 inches to 22 inches).
181
8. Outside diameter of disks = 12 inches
Fig.C21 Two-disk rotor model with increased outside diameter of disks
Fig.C21 shows a two-disk rotor model with the diameters of the disks increased
from 9 inches to 12 inches. The thickness of the aluminum sleeve is 0.5 inches, the same
as in the baseline model. The remaining geometry and material properties are the same
as the baseline model. The internal friction moments are applied at one sleeve-disk
interface with cross-coupling, whereas the other interface has internal moment
coefficients with no cross-coupling. The simulations using the XLTRCTM show that the
182
rotor is more unstable as compared to the baseline rotor model. The first critical speed of
the rotor is 5700 rpm. The threshold speed of instability of the rotor is 7000 rpm at 5650
cpm. The increased instability of the rotor model as compared to the baseline model can
be explained by increase in the diameter of the disks. As the disk diameter is increased,
the bending stiffness of the aluminum sleeve increases. Due to increase in bending
stiffness, larger slip is developed at the sleeve-disk interface because the sleeve is much
more stiff as compared to the steel shaft, that now carries a larger weight of the
disks(due to increase in disk diameters). The larger slip gives rise to the larger internal
friction moments, resulting in higher instability.
9. Outside diameter of disks = 20 inches
Fig.C22 Two-disk rotor model with increased outside diameter of disks
183
Fig.C22 shows a two-disk rotor model with outside diameter of the disks as 20
inches, whereas the thickness of the sleeve is 0.5 inches as in the baseline model. The
simulations using the XLTRCTM software show that the first critical speed of the rotor is
4000 rpm and the threshold speed of instability of the rotor is 6000 rpm. The increase of
instability is explained by the larger bending stiffness of the sleeve and a larger slip
developed at the shrink fit interface of the disk and the sleeve.
10. Disk axial width = 3 inches
Fig.C23 Two-disk rotor model with disk axial width = 3 inches
184
Fig.C23 shows a two-disk rotor model with the axial width of the disks increased
to 3 inches as compared to the baseline rotor model in which the axial width is 2 inches.
The internal moments distribution is the same as in the baseline model. In the rotor
model shown in Fig.C23, the axial contact between the sleeve and the disk at the
interface where cross-coupled moment coefficients are applied is 2 inches. In the
baseline rotor model, the axial contact length between the sleeve and the disk at the
friction interface is 2 inches. The rotordynamic simulations using the XLTRCTM
software show that the the first critical speed of the rotor is 6040 rpm and the threshold
speed of instability is 11500 rpm at a frequency 5927 cpm (unstable mode frequency).
The increased stability of the rotor can be explained due to larger axial contact length at
the friction interface. The larger the contact length is, the lesser is the amount of relative
slip at the interface. A lesser amount of slip at the interface will result in smaller
magnitude of the internal moments and consequently higher stability of the rotor.
185
APPENDIX D
FOUNDATION MODAL PARAMETERS AND BALL BEARING
PARAMETERS
As described earlier, modal parameters of the foundation are estimated using
impact and shaker tests on the foundation, with the rotor removed from the bearings
while the measurements are conducted. In this way, only the foundation parameters are
determined, which can then be inputted to a separate file linked to XLTRCTM for
performing the simulations of the rotor-bearing system using XLTRCTM.
Both impact tests and shaker tests were performed in the horizontal direction, to
check the results of those measurements against each other. In the vertical direction, only
impact hammer tests were performed, since a shaker can not be used in this direction.
The impact hammer tests were conducted in the horizontal direction. It is usually
the simplest method of modal testing requiring only an impact hammer and an
accelerometer, along with a dynamic signal analyzer. The results are shown on the next
page:
186
Fig. D1 Impact tests in horizontal direction
The shaker tests were also conducted in the horizontal direction to provide a
check for the measurements using impact tests. A shaker was connected to the bearing
housing using a small plate. On the opposite side of the bearing housing was placed a
high sensitivity accelerometer. The data was collected on a dynamic signal analyzer and
the results of acceleration to force ratio using several tests in horizontal direction in
frequency domain are shown on the next page:
M agnitude A/F
0.00E+002.00E+004.00E+006.00E+008.00E+001.00E+011.20E+011.40E+01
0 50 100 150 200Frequency (Hz)
A/F
M agnitude (A/F)
0.00E+002.00E+004.00E+006.00E+008.00E+001.00E+011.20E+011.40E+011.60E+01
0 50 100 150 200Frequency (Hz)
A/F
187
M agnitude Plot(A/F)
0.00E+002.00E+004.00E+006.00E+008.00E+001.00E+011.20E+011.40E+011.60E+01
0 50 100 150 200Frequency (Hz)
A/F
M agnitude(A/F)
0.00E+002.00E+00
4.00E+006.00E+008.00E+001.00E+01
1.20E+011.40E+01
0 50 100 150 200Frequency(Hz )
A/F
Fig. D2 Shaker test results for foundation in horizontal direction
As can be seen from the experimental data presented in Fig.D2 for two separate
tests performed using the shaker, the horizontal natural frequency is around 125 Hz.
In a similar way, the impact hammer tests were conducted in the vertical
direction and the following results for the ratio of acceleration to force are shown:
188
M agnitude A/F
0.00E+00
1.00E-01
2.00E-01
3.00E-014.00E-01
5.00E-01
6.00E-01
7.00E-01
8.00E-01
0 50 100 150 200Frequency (Hz )
A/F
M agnitude of A/F
0.00E+001.00E-012.00E-013.00E-014.00E-015.00E-016.00E-017.00E-018.00E-019.00E-01
0 50 100 150 200 250Frequency (Hz)
A/F
Fig. D3 Impact tests in vertical direction
From Fig.D3, it can be seen that for vertical direction also, the first natural
frequency is around 125 Hz. That the two natural frequencies in the horizontal and the
vertical direction are almost identical is a coincidence (not a planned result).
189
In order to get an estimate of the numerical values of the modal parameters in
both the horizontal and the vertical direction, it is instructive first to consider the
following single degree of freedom system subjected to harmonic excitation, as shown in
Fig.D4 below:
Fig. D4 Single degree of freedom system subjected to harmonic excitation
The differential equation of motion for the above simple system can be written as
follows:
)sin()()()()( 0 tFtFtKytyCtyM ω==++•••
(D1)
C K
M
y
F(t)
190
In equation (1), the coefficients on left hand side for acceleration, velocity and
displacement are mass, damping and stiffness, respectively. On the right hand side, there
is a time dependent force, with ‘ω’ as the frequency of excitation.
Using the methods of complex variables that are employed to solve differential
equations as (1), the following ratio for the acceleration amplitude of the mass to the
amplitude of excitation force can be established:
222
2
)()( ωω
ω
CMKFY
+−=
••
(D2)
Based on equation (2), it can be seen that if a system is simplified as a single
degree of freedom system and considering linear viscous damping, then acceleration to
force ratio is dependent on the system modal parameters K, C, and M.
Equation (D2) is utilized to obtain the modal parameter values of the foundation
in horizontal and the vertical directions by curve fitting equation (D2) to the shaker test
data. This curve fitting is performed in Microsoft Excel and the modal parameters for the
two directions are described in Table D1:
191
Table D1. Modal parameters of foundation in two directions
The foundation’s first natural frequency in the horizontal direction as can be seen
from the measurement results is around 125 Hz. Before the stiffeners were installed, the
natural frequency in the horizontal direction was 70 Hz [18]. Thus, installation of the
structures substantially enhanced the stiffness of the foundation in horizontal direction.
This can be seen in the following table:
Table D2. Comparison of modal parameters in two directions
Modal Parameters Horizontal Direction Vertical Direction
Stiffness (lb/in) 295000 3700000
Mass (lb-s2/in) 0.48 6.4
Damping (lb-s/in) 65.5 1350
Horizontal direction Vertical direction Modal Parameters
Previous Current Previous Current
Stiffness (lb/in) 90590 295000 NA 3700000
Mass(lb-s2/in) 0.5 0.48 NA 6.4
Damping (lb-s/in) 45.5 65.5 NA 1350
192
ADDITION OF THIRD STIFFENER STRUCTURE
A third stiffener structure was added to the foundation for the experimental
testing of the two-disk rotor configurations. The third structure was added to stiffen up
the foundation further in the horizontal direction and to reduce the stiffness asymmetry
further, because this is one of the important factors that causes rotordynamic instability
due to slippage in shrink fit and interference fit joints.
The same procedure as outlined in the previous pages is applied to determine
experimentally the modal parameters of the foundation. The experimental results and the
values of the foundation modal parameters in the horizontal direction are described
below:
Foundation left bearing- Horizontal direction
0
1
2
3
4
5
6
7
8
0 100 200 300 400
Frequency - Hz
Tran
sfer
func
tion
(A/F
)
Experimental Curve Fit
Fig.D5 Acceleration-Force transfer function graph of the foundation in horizontal direction
193
From the curve fitting of the acceleration /force data in the frequency domain, the
following foundation modal parameters are found to best fit the experimental data:
Table D3. Foundation modal parameters after the addition of stiffener structure
Table D3 shows that the addition of the third stiffener structure results in an
increase of foundation’s horizontal stiffness, mass and damping.
BALL BEARING PARAMETERS
The stiffness and damping parameters of the ball bearings are presented below.
These parameters are calculated by specifying the bearing parameters such as the
number of balls, bearing inside and outside diameters and the axial width of the bearing.
The built-in code in XLTRCTM calculates the corresponding bearing stiffness and
damping coefficients, which can be used to perform rotordynamic simulations by
connecting the ball bearing coefficients file to the ground or foundation, which is the
procedure adopted in the rotordynamic simulations in this Dissertation.
The ball bearing coefficients are presented in Table D4:
Modal Parameters Horizontal Direction Vertical Direction
Stiffness (lb/in) 48000 3700000
Mass (lb-s2/in) 0.835 6.4
Damping (lb-s/in) 112 1350
194
Table D4. Ball bearing coefficients
Table D4 shows that the ball bearing direct stiffness coefficients vary slightly
with the rotational speed and they are of the order of magnitude of 1e+5 lb/in. There is
no cross-coupling and no direct damping offered through the ball bearings.
Speed Kxx Kxy Kyx Kyy Cxx Cxy Cyx Cyy
RPM Lb/in Lb/in Lb/in Lb/in Lb-s/in Lb-s/in Lb-s/in Lb-s/in
0 5.6e+5 0 0 5.6e+5 0 0 0 0
2000 5.7e+5 0 0 5.7e+5 0 0 0 0
4000 5.3e+5 0 0 5.3e+5 0 0 0 0
6000 4.8e+5 0 0 4.8e+5 0 0 0 0
8000 4.5e+5 0 0 4.5e+5 0 0 0 0
10000 4.3e+5 0 0 4.3e+5 0 0 0 0
195
APPENDIX E
SUMMARY OF DISSERTATION
On the recommendation of the Committee members, the final examination
presentation slides given by the student in the defense of his doctorate dissertation are
presented in its entire details in this Appendix. The purpose of presenting these slides is
to provide the readers a quick, graphical summary of the student’s doctoral research.
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
VITA
Syed Muhammad Mohsin Jafri received his Bachelor of Engineering degree from
NED University of Engineering and Technology, Karachi, Pakistan in May 1999. He
enrolled in the Mechanical Engineering Department at Texas A&M University, College
Station in August 2000 to pursue a Master of Science degree, and received the degree in
May 2004. He continued his education to pursue a Doctor of Philosophy degree in
Mechanical Engineering at Texas A&M University, College Station. His research and
educational background during the graduate studies include dynamics, vibrations and
rotordynamics.
Mr. Jafri can be reached at Deep Sea Engineering & Management, Inc., 10333
Richmond Avenue, Suite 250, Houston, TX 77042. His e-mail address is